@devries-cameron20 / composio Mirror · read-only / commits / f19794cb8
Fork 0

Composio powers 1000+ toolkits, tool search, context management, authentication, and a sandboxed workbench to help you build AI agents that turn intent into action.

Mirrored from https://github.com/Camo67/composio.git

Synced Nov 21, 04:45 UTC

docs: centralize the 1,500+ apps/toolkits count (#5400)

> Written by an AI agent on behalf of @olearycrew, following their spec. @olearycrew has not reviewed it yet. ## Summary The docs advertised several catalog sizes, including `500+ apps`, `1000+ tools`, and `1,000+ apps`. Use the requested `1,500+` from one shared constant, and name what it counts consistently: - Marketing copy, including site descriptions and general product introductions, says **1,500+ apps**. - Technical documentation about the catalog says **1,500+ toolkits**. - **Tools** means the actions within an individual toolkit. It is never the name for the catalog count. This follows Brendan's request and wording rule. Designated reviewer: @jkomyno. Please review the full diff before merging. ## Changes - Set the existing `TOOLKIT_COUNT_LABEL` in `docs/lib/toolkit-count.ts` to `1,500+`, replacing the previously derived value. - Resolve `%TOOLKIT_COUNT%` through the existing Fumadocs build hooks and frontmatter schema. The same resolver covers raw Markdown readers, search records, and the generated docs-agent snapshot. No dependencies added. - Convert the agent's Markdown context to Eve's supported TypeScript instructions module so it imports the constant. Its resolved prompt is unchanged apart from the count. - Update the meta-tools generator and regenerate the overview and search-tool page from the checked-in API snapshot. The generated source keeps the placeholder, and the search-tool description uses our maintained copy instead of the outdated API claim. - Keep homepage copy, benefit descriptions, and general `llms` introductions in apps. Use toolkits in reference and meta-tool documentation. The custom-provider guide now says “call tools across 1,500+ toolkits”. - Replace Xero's cross-catalog “1000+ tools” claim with “schemas from every toolkit” in the editorial article, then regenerate the published page. - Record the terminology rule in `docs/AGENTS.md` and verify it in the existing count, metadata, and rendered-output tests. - Generate the agent snapshot in `pretest`, before static tests import it. This fixes the fresh-checkout `Cannot find module './docs-index'` error from the count test. - Include the connected-account troubleshooting guide in the Platform authentication sidebar. The navigation coverage assertion remains intact and now passes. - Extract rendered paragraph text with Bun's native `HTMLRewriter`, following the existing provider integration-test pattern. No hand-written HTML filtering or new dependency. Regression cases cover paragraph-like markup inside scripts and both script closing-tag forms. ## Type of change - [ ] Bug fix - [ ] New feature - [x] Refactor/Chore - [x] Documentation - [ ] Breaking change ## How Has This Been Tested? On the final revision, from `docs/`: - Reproduced the CI errors before the fix. With the ignored agent snapshot absent, the count test failed to import `./docs-index`. The navigation suite separately failed because `/docs/authentication/session-cant-find-account` was omitted from the Platform sidebar. - Removed the local generated snapshot, then ran the same `bun run test` command CI runs. The updated `pretest` generated it and the full static suite passed: **835 passed, 0 failed**. - `GITHUB_ACTIONS=true bun run build` passed, including production TypeScript checking. The flag selects the existing two-worker configuration. ```text ✓ Compiled successfully in 21.7s Finished TypeScript in 8.1s ... ✓ Generating static pages using 2 workers (2600/2600) in 15.0s ``` - Started the production server with `bun run start --hostname 127.0.0.1 --port 3047`, then ran `TEST_BASE_URL=http://127.0.0.1:3047 bun run test:integration`: **245 passed, 1 skipped, 0 failed**. The existing production-only fallback test requires `COMPOSIO_API_KEY`. The count tests now use the HTML parser to check actual paragraphs rather than a metadata or hydration match. - The new parser regression passes for both `</script>` and `</script >` and ignores fake paragraph markup inside script contents. - `bun run lint` passed with existing warnings. `bun run check:kb-semantic --allow-stale` passed with the expected stale-artifact warning and keyword fallback. The existing rebuild workflow owns the generated embeddings artifact. - `git diff --check` passed. The earlier root agent-guidance validation also passed; no guidance changed in this upkeep commit. Local tools: Bun 1.4.2 and Node 24.22.1. CI uses the repository pins. ## Screenshots (if applicable) Copy and metadata were checked in the built HTTP responses. No layout changes or browser screenshots. | Surface | Before on `origin/next` | Final built response | | --- | --- | --- | | Homepage and social-card copy | Derived count, mixed apps/tools wording | `1,500+ apps` | | OpenAI benefit description | `use 1000+ tools` | `use 1,500+ apps` | | `/reference` and `/reference.md` | `1000+ toolkits` | `1,500+ toolkits` | | Meta-tools overview | `500+ apps` | `1,500+ toolkits` | | Custom-provider guide | `call Composio's 1000+ tools` | `call tools across 1,500+ toolkits` | | Claude Code plugin description | `Act on 1,000+ apps` | `Act on 1,500+ apps` | ## Checklist - [x] I have read the Code of Conduct and this PR adheres to it - [ ] Human author has reviewed every line of the diff. The agent has reviewed it; full human review is pending. - [x] Any third-party docs links or promotional content follow the third-party links policy. No links added. - [x] Static tests, build, and production integration tests passed. See the one credential-dependent integration skip above. - [x] I updated documentation as needed - [x] I added tests for count resolution, generation, and rendered terminology - [x] Changeset is not applicable to this docs-only change ## Additional context ### Open scope question Historical entries remain unchanged as originally requested: `content/changelog/03-25-26-mcp-endpoint-removed.mdx`, `decisions/examples.md`, and `decisions/cookbooks-revamp-plan.md`. Confirm they should continue to describe their original point in time. The apps/toolkits/tools wording questions are resolved by Brendan's rule above. ### Generated snapshots Raw `public/data/meta-tools.json` retains the upstream API payload; rendered docs use the count-aware override. `kb/source/toolkits/xero/public.md` is the imported upstream snapshot; the editorial article owns the revised published prose. `kb/semantic-index.json` was not hand-edited. Its old content hashes fail validation after content changes, causing keyword fallback until rebuilt. The existing “Docs - Rebuild KB Semantic Artifact” workflow owns that rebuild after the PR's Docs - Tests run; the required API key is unavailable locally. ### ELI5 The catalog size was copied into many pages, so the numbers drifted apart. Pages now fill in one shared number and use “apps” for the product promise and “toolkits” when explaining the technical catalog. <details> <summary>Every touched file (47)</summary> - `docs/AGENTS.md` (updated). - `docs/agent/instructions/context.md` (removed; replaced by the TypeScript instructions module). - `docs/agent/instructions/context.ts` (added). - `docs/agent/lib/docs.ts` (updated). - `docs/app/layout.tsx` (updated). - `docs/app/llms-full.txt/route.ts` (updated). - `docs/app/llms-index.txt/route.ts` (updated). - `docs/app/llms.txt/route.ts` (updated). - `docs/components/home-features.tsx` (updated). - `docs/content/docs/claude-code-plugin.mdx` (updated). - `docs/content/docs/composio-connect.mdx` (updated). - `docs/content/docs/migration-guide/mcp-servers-to-sessions.mdx` (updated). - `docs/content/docs/providers/anthropic.mdx` (updated). - `docs/content/docs/providers/autogen.mdx` (updated). - `docs/content/docs/providers/crewai.mdx` (updated). - `docs/content/docs/providers/custom-providers/index.mdx` (updated). - `docs/content/docs/providers/custom-providers/python.mdx` (updated). - `docs/content/docs/providers/custom-providers/typescript.mdx` (updated). - `docs/content/docs/providers/google.mdx` (updated). - `docs/content/docs/providers/langchain.mdx` (updated). - `docs/content/docs/providers/llamaindex.mdx` (updated). - `docs/content/docs/providers/mastra.mdx` (updated). - `docs/content/docs/providers/openai.mdx` (updated). - `docs/content/docs/providers/vercel.mdx` (updated). - `docs/content/docs/quickstart.mdx` (updated). - `docs/content/examples/general-agent-with-pi.mdx` (updated). - `docs/content/examples/harness-integration.mdx` (updated). - `docs/content/examples/imessage-agent.mdx` (updated). - `docs/content/examples/local-sandbox-pr-reviewer.mdx` (updated). - `docs/content/kb/guide/toolkits-xero.mdx` (updated). - `docs/content/reference/index.mdx` (updated). - `docs/content/reference/v3/index.mdx` (updated). - `docs/content/toolkits/meta-tools/index.mdx` (updated). - `docs/content/toolkits/meta-tools/search_tools.mdx` (updated). - `docs/kb/articles/toolkits-xero.md` (updated). - `docs/lib/home-navigation.ts` (updated). - `docs/lib/meta-tool-overrides.ts` (updated). - `docs/lib/search-index.ts` (updated). - `docs/lib/source.ts` (updated). - `docs/lib/toolkit-count.ts` (updated). - `docs/package.json` (updated). - `docs/scripts/build-agent-index.ts` (updated). - `docs/scripts/generate-meta-tools.ts` (updated). - `docs/source.config.ts` (updated). - `docs/tests/integration/toolkit-count.test.ts` (added). - `docs/tests/static/product-navigation.test.ts` (updated). - `docs/tests/static/toolkit-count.test.ts` (added). </details>

Brendan O'Leary committed Nov 16, 2026 at 01:59 UTC f19794cb8f5b844fa310d1167558fd31c62b4116
47 files changed +237 -121
docs/AGENTS.md
+2
@@ -41,6 +41,8 @@ bun run inventory:code-examples
41
42 ## Rules
43
44 +- Catalog-size copy uses "apps" in marketing text, including site descriptions, "toolkits" in technical documentation about toolkits, and "tools" only for an individual toolkit's actions. Use `TOOLKIT_COUNT_LABEL` from `lib/toolkit-count.ts` or `%TOOLKIT_COUNT%` in Markdown/MDX for the shared count.
45 +
46 - Production Twoslash and `check:typescript-snippets` cover different maintained TypeScript surfaces. Read `docs/agent-guidance/context/twoslash.md` for coverage, installed-version boundaries, and annotation rules.
47 - Maintained operation examples offer Python, TypeScript, and API options, or a specific justified exception. Read `docs/agent-guidance/context/code-examples.md` for the full inventory, generated/historical limits, check commands, and scoped markers. Passing checks do not establish semantic equivalence or justify an exemption.
48 - Parse untyped or external data (JSON files, fetched payloads, framework page data) once at the boundary with zod schemas and let `z.infer` types flow downstream. Never hand-roll structural guards (`'x' in obj` / `typeof` chains), cast parsed JSON with `as`, or fake validation with `z.custom(() => true)`.
docs/agent/instructions/context.md deleted
-61
@@ -1,61 +0,0 @@
1 -# Composio concept map
2 -
3 -This is your always-on map of Composio's concepts and the canonical page for each. Use it to ground answers and to link the right page. Prefer these links over anything `search_docs` returns. Use the full bounded content from `search_docs` first, and call `read_doc` on the relevant page only when you need detail beyond that included content.
4 -
5 -## Core model
6 -
7 -- **Session** — the runtime context for one user. `composio.create(userId)` (or the canonical `composio.sessions.create(...)`) returns a session that ties together the user, toolkits, authentication, connected accounts, and a code-execution sandbox. By default it exposes meta tools the agent calls at runtime to discover, authenticate, and execute tools. → [What is a session?](/docs/how-composio-works)
8 -- **Configuring a session** — filter toolkits and tools, set auth configs, select connected accounts, preload tools, the direct-tools preset, sandbox tier, and session methods (`tools()`, `toolkits()`, `authorize()`). → [Configuring sessions](/docs/configuring-sessions)
9 -- **Sessions via MCP** — create a session with `{ mcp: true }` to expose `session.mcp.url` / `session.mcp.headers` for any MCP client. → [Using sessions via MCP](/docs/sessions-via-mcp)
10 -- **Reusing sessions** — sessions persist on the server; store the session ID and reuse with `composio.use(sessionId)`, or update in place with `session.update(...)`. → [What is a session?](/docs/how-composio-works)
11 -
12 -## Tools and toolkits
13 -
14 -- **Toolkit** — a collection of related tools for a service (e.g. `github`, `gmail`). A **tool** is one action, named `{TOOLKIT}_{ACTION}` (e.g. `GITHUB_CREATE_ISSUE`). Every toolkit is discoverable by default; restrict with the `toolkits` config. → [Configuring sessions](/docs/configuring-sessions)
15 -- **Meta tools** — the fixed set every session exposes (`COMPOSIO_SEARCH_TOOLS`, `COMPOSIO_GET_TOOL_SCHEMAS`, `COMPOSIO_MANAGE_CONNECTIONS`, `COMPOSIO_MULTI_EXECUTE_TOOL`, `COMPOSIO_REMOTE_WORKBENCH`, `COMPOSIO_REMOTE_BASH_TOOL`). → [Meta tools reference](/toolkits/meta-tools)
16 -- **Providers** — adapter packages that format Composio tools for a framework (OpenAI, Anthropic, Vercel AI SDK, LangChain, Mastra, Pi, …). → [Providers](/docs/providers)
17 -- **Is a toolkit / integration supported?** Composio has 1000+ toolkits, and `search_docs` indexes the catalog. A matching `/toolkits/<slug>` result means **yes, it's supported** — answer from the returned content when sufficient, call `read_doc` only for more details, and point the user at [the toolkits directory](/toolkits). If nothing matches, it isn't a built-in toolkit; suggest [proxy execute](/docs/extending-sessions/proxy-execute) or a [local tool](/docs/extending-sessions/local-tools-and-toolkits) for an API you already have.
18 -
19 -## Authentication
20 -
21 -- **How auth works** — Composio uses Connect Links (hosted auth pages) and **auth configs** (per-toolkit blueprints) to create **connected accounts** (stored credentials) scoped to a userID. This is the page for "how does authentication work". → [Authentication](/docs/authentication)
22 -- **Auth schemes / modes** — a toolkit's auth config uses one of `OAUTH2`, `API_KEY`, `BEARER_TOKEN`, or `BASIC`. When you mention a toolkit's auth mode, link the reference rather than explaining it inline. → [Auth schemes](/reference/api-reference/auth-configs#auth-schemes) for what each mode is and when it's used, [Authentication](/docs/authentication) for how auth works overall, and [Managed vs custom auth](/docs/authentication/custom-app-vs-managed-app) for choosing or bringing your own scheme.
23 -- **In-chat auth** — the agent prompts the user to connect via `COMPOSIO_MANAGE_CONNECTIONS`. → [In-chat authentication](/docs/authentication#in-chat-authentication)
24 -- **Manual auth** — generate Connect Links yourself with `session.authorize()`. → [Manually authenticating users](/docs/authentication/manually-authenticating)
25 -- **Managed vs custom auth** — use Composio's managed OAuth apps, or bring your own for branding/scopes. → [Managed vs custom auth](/docs/authentication/custom-app-vs-managed-app)
26 -- **White-labeling** — remove Composio branding from the auth flow. → [White-labeling authentication](/docs/authentication/white-labeling-authentication)
27 -- **Callback identity verification** — opt-in per-project defense against OAuth session fixation, documented in the connected-accounts API overview: set a verifier URL and Composio holds each OAuth connection until the developer's server confirms the returning user (`complete_auth`); once set it covers every OAuth connection in the project. This is the section for "connections from the dashboard stopped completing" and "a real `user_id` is now required". → [Callback identity verification](/reference/api-reference/connected-accounts#callback-identity-verification)
28 -- **Importing existing connections** — pass in API keys or bearer tokens you already hold. → [Importing existing connections](/docs/authentication/importing-existing-connections)
29 -- **Multiple accounts per user** — e.g. work and personal Gmail. → [Managing multiple connected accounts](/docs/authentication/managing-multiple-connected-accounts)
30 -- **Shared connections** — share one connected account across users via an ACL. → [Shared connections](/docs/extending-sessions/shared-connections)
31 -
32 -## Triggers and webhooks
33 -
34 -- **Triggers** — receive structured payloads when something happens in a connected app (webhook or polling). → [Triggers](/docs/triggers)
35 -- **Setting up triggers** — create, manage, and subscribe to trigger events. → [Creating triggers](/docs/setting-up-triggers/creating-triggers), [Subscribing to events](/docs/setting-up-triggers/subscribing-to-events), [Managing triggers](/docs/setting-up-triggers/managing-triggers)
36 -- **Webhook verification** — verify inbound webhook signatures. → [Webhook verification](/docs/webhook-verification)
37 -
38 -## Extending sessions
39 -
40 -- **Sandbox** (previously "workbench") — a persistent Python environment at `/mnt/files/` for bulk operations and large responses; files via `session.experimental.files`. → [Sandbox](/docs/sandbox)
41 -- **Local tools and toolkits** — define in-process tools that run alongside Composio tools. → [Local tools and toolkits](/docs/extending-sessions/local-tools-and-toolkits)
42 -- **Proxy execute** — call any toolkit HTTP endpoint with `session.proxyExecute(...)` and let Composio inject auth. → [Proxy execute](/docs/extending-sessions/proxy-execute)
43 -
44 -## Platform API (reference only)
45 -
46 -These features are documented **only in the API reference** — there is no `/docs` guide. When asked how to do them programmatically, read and link the reference page (don't say it isn't documented).
47 -
48 -- **Projects** — Composio's multi-tenancy primitive. Inside an organization, projects are isolated environments that scope API keys, connected accounts, auth configs, and webhooks. Create, list, update, delete, and regenerate a project's API key via the API (using your **organization API key**). This is the page for "how do I programmatically create projects". → [Projects](/reference/api-reference/projects)
49 -- **Logs** — individual tool-execution events (one record per call) for debugging and tracing. → [Logs](/reference/api-reference/logs)
50 -- **Files** — files tools read and write during execution, exchanged via presigned URLs. → [Files](/reference/api-reference/files)
51 -
52 -## Getting started and reference
53 -
54 -- **Quickstart** → [Quickstart](/docs/quickstart)
55 -- **Glossary** → [Glossary](/reference/glossary)
56 -- **SDK reference (TypeScript Session)** → [Session](/reference/sdk-reference/typescript/session)
57 -- **API reference** → [API reference](/reference/api-reference)
58 -
59 -## Legacy — do not lead with these
60 -
61 -Direct tool execution and the `tools-direct/*` pages are the **legacy**, pre-session API. Do not mention "direct execution" or link these pages unless the user explicitly asks about the low-level / direct-execution API. For everything else, answer with the session-based model above.
docs/agent/instructions/context.ts new
+67
@@ -0,0 +1,67 @@
1 +import { defineInstructions } from 'eve/instructions';
2 +import { TOOLKIT_COUNT_LABEL } from '../../lib/toolkit-count';
3 +
4 +export default defineInstructions({
5 + markdown: `# Composio concept map
6 +
7 +This is your always-on map of Composio's concepts and the canonical page for each. Use it to ground answers and to link the right page. Prefer these links over anything \`search_docs\` returns. Use the full bounded content from \`search_docs\` first, and call \`read_doc\` on the relevant page only when you need detail beyond that included content.
8 +
9 +## Core model
10 +
11 +- **Session** — the runtime context for one user. \`composio.create(userId)\` (or the canonical \`composio.sessions.create(...)\`) returns a session that ties together the user, toolkits, authentication, connected accounts, and a code-execution sandbox. By default it exposes meta tools the agent calls at runtime to discover, authenticate, and execute tools. → [What is a session?](/docs/how-composio-works)
12 +- **Configuring a session** — filter toolkits and tools, set auth configs, select connected accounts, preload tools, the direct-tools preset, sandbox tier, and session methods (\`tools()\`, \`toolkits()\`, \`authorize()\`). → [Configuring sessions](/docs/configuring-sessions)
13 +- **Sessions via MCP** — create a session with \`{ mcp: true }\` to expose \`session.mcp.url\` / \`session.mcp.headers\` for any MCP client. → [Using sessions via MCP](/docs/sessions-via-mcp)
14 +- **Reusing sessions** — sessions persist on the server; store the session ID and reuse with \`composio.use(sessionId)\`, or update in place with \`session.update(...)\`. → [What is a session?](/docs/how-composio-works)
15 +
16 +## Tools and toolkits
17 +
18 +- **Toolkit** — a collection of related tools for a service (e.g. \`github\`, \`gmail\`). A **tool** is one action, named \`{TOOLKIT}_{ACTION}\` (e.g. \`GITHUB_CREATE_ISSUE\`). Every toolkit is discoverable by default; restrict with the \`toolkits\` config. → [Configuring sessions](/docs/configuring-sessions)
19 +- **Meta tools** — the fixed set every session exposes (\`COMPOSIO_SEARCH_TOOLS\`, \`COMPOSIO_GET_TOOL_SCHEMAS\`, \`COMPOSIO_MANAGE_CONNECTIONS\`, \`COMPOSIO_MULTI_EXECUTE_TOOL\`, \`COMPOSIO_REMOTE_WORKBENCH\`, \`COMPOSIO_REMOTE_BASH_TOOL\`). → [Meta tools reference](/toolkits/meta-tools)
20 +- **Providers** — adapter packages that format Composio tools for a framework (OpenAI, Anthropic, Vercel AI SDK, LangChain, Mastra, Pi, …). → [Providers](/docs/providers)
21 +- **Is a toolkit / integration supported?** Composio has ${TOOLKIT_COUNT_LABEL} toolkits, and \`search_docs\` indexes the catalog. A matching \`/toolkits/<slug>\` result means **yes, it's supported** — answer from the returned content when sufficient, call \`read_doc\` only for more details, and point the user at [the toolkits directory](/toolkits). If nothing matches, it isn't a built-in toolkit; suggest [proxy execute](/docs/extending-sessions/proxy-execute) or a [local tool](/docs/extending-sessions/local-tools-and-toolkits) for an API you already have.
22 +
23 +## Authentication
24 +
25 +- **How auth works** — Composio uses Connect Links (hosted auth pages) and **auth configs** (per-toolkit blueprints) to create **connected accounts** (stored credentials) scoped to a userID. This is the page for "how does authentication work". → [Authentication](/docs/authentication)
26 +- **Auth schemes / modes** — a toolkit's auth config uses one of \`OAUTH2\`, \`API_KEY\`, \`BEARER_TOKEN\`, or \`BASIC\`. When you mention a toolkit's auth mode, link the reference rather than explaining it inline. → [Auth schemes](/reference/api-reference/auth-configs#auth-schemes) for what each mode is and when it's used, [Authentication](/docs/authentication) for how auth works overall, and [Managed vs custom auth](/docs/authentication/custom-app-vs-managed-app) for choosing or bringing your own scheme.
27 +- **In-chat auth** — the agent prompts the user to connect via \`COMPOSIO_MANAGE_CONNECTIONS\`. → [In-chat authentication](/docs/authentication#in-chat-authentication)
28 +- **Manual auth** — generate Connect Links yourself with \`session.authorize()\`. → [Manually authenticating users](/docs/authentication/manually-authenticating)
29 +- **Managed vs custom auth** — use Composio's managed OAuth apps, or bring your own for branding/scopes. → [Managed vs custom auth](/docs/authentication/custom-app-vs-managed-app)
30 +- **White-labeling** — remove Composio branding from the auth flow. → [White-labeling authentication](/docs/authentication/white-labeling-authentication)
31 +- **Callback identity verification** — opt-in per-project defense against OAuth session fixation, documented in the connected-accounts API overview: set a verifier URL and Composio holds each OAuth connection until the developer's server confirms the returning user (\`complete_auth\`); once set it covers every OAuth connection in the project. This is the section for "connections from the dashboard stopped completing" and "a real \`user_id\` is now required". → [Callback identity verification](/reference/api-reference/connected-accounts#callback-identity-verification)
32 +- **Importing existing connections** — pass in API keys or bearer tokens you already hold. → [Importing existing connections](/docs/authentication/importing-existing-connections)
33 +- **Multiple accounts per user** — e.g. work and personal Gmail. → [Managing multiple connected accounts](/docs/authentication/managing-multiple-connected-accounts)
34 +- **Shared connections** — share one connected account across users via an ACL. → [Shared connections](/docs/extending-sessions/shared-connections)
35 +
36 +## Triggers and webhooks
37 +
38 +- **Triggers** — receive structured payloads when something happens in a connected app (webhook or polling). → [Triggers](/docs/triggers)
39 +- **Setting up triggers** — create, manage, and subscribe to trigger events. → [Creating triggers](/docs/setting-up-triggers/creating-triggers), [Subscribing to events](/docs/setting-up-triggers/subscribing-to-events), [Managing triggers](/docs/setting-up-triggers/managing-triggers)
40 +- **Webhook verification** — verify inbound webhook signatures. → [Webhook verification](/docs/webhook-verification)
41 +
42 +## Extending sessions
43 +
44 +- **Sandbox** (previously "workbench") — a persistent Python environment at \`/mnt/files/\` for bulk operations and large responses; files via \`session.experimental.files\`. → [Sandbox](/docs/sandbox)
45 +- **Local tools and toolkits** — define in-process tools that run alongside Composio tools. → [Local tools and toolkits](/docs/extending-sessions/local-tools-and-toolkits)
46 +- **Proxy execute** — call any toolkit HTTP endpoint with \`session.proxyExecute(...)\` and let Composio inject auth. → [Proxy execute](/docs/extending-sessions/proxy-execute)
47 +
48 +## Platform API (reference only)
49 +
50 +These features are documented **only in the API reference** — there is no \`/docs\` guide. When asked how to do them programmatically, read and link the reference page (don't say it isn't documented).
51 +
52 +- **Projects** — Composio's multi-tenancy primitive. Inside an organization, projects are isolated environments that scope API keys, connected accounts, auth configs, and webhooks. Create, list, update, delete, and regenerate a project's API key via the API (using your **organization API key**). This is the page for "how do I programmatically create projects". → [Projects](/reference/api-reference/projects)
53 +- **Logs** — individual tool-execution events (one record per call) for debugging and tracing. → [Logs](/reference/api-reference/logs)
54 +- **Files** — files tools read and write during execution, exchanged via presigned URLs. → [Files](/reference/api-reference/files)
55 +
56 +## Getting started and reference
57 +
58 +- **Quickstart** → [Quickstart](/docs/quickstart)
59 +- **Glossary** → [Glossary](/reference/glossary)
60 +- **SDK reference (TypeScript Session)** → [Session](/reference/sdk-reference/typescript/session)
61 +- **API reference** → [API reference](/reference/api-reference)
62 +
63 +## Legacy — do not lead with these
64 +
65 +Direct tool execution and the \`tools-direct/*\` pages are the **legacy**, pre-session API. Do not mention "direct execution" or link these pages unless the user explicitly asks about the low-level / direct-execution API. For everything else, answer with the session-based model above.
66 +`,
67 +});
docs/agent/lib/docs.ts
+3 -2
@@ -6,6 +6,7 @@
6 * search tool can heavily downrank them.
7 */
8
9 +import { resolveToolkitCount } from '../../lib/toolkit-count';
10 import { existsSync, readdirSync, readFileSync } from 'node:fs';
11 import { dirname, join, relative } from 'node:path';
12 import { fileURLToPath } from 'node:url';
@@ -434,7 +435,7 @@ export function buildIndex(): DocPage[] {
435 if (!url) continue;
436 let raw: string;
437 try {
437 - raw = readFileSync(absPath, 'utf8');
438 + raw = resolveToolkitCount(readFileSync(absPath, 'utf8'));
439 } catch {
440 continue;
441 }
@@ -507,7 +508,7 @@ export function readPageByUrl(url: string): { title: string; raw: string } | und
508 ];
509 for (const candidate of candidates) {
510 try {
510 - const raw = readFileSync(candidate, 'utf8');
511 + const raw = resolveToolkitCount(readFileSync(candidate, 'utf8'));
512 const { title } = parseFrontmatter(raw);
513 return { title: title || url, raw };
514 } catch {
docs/app/layout.tsx
+1 -1
@@ -34,7 +34,7 @@ const defaultLinks = defaultLinkSlugs.flatMap(({ slug, source: pageSource }) =>
34 return [{ title: page.data.title, description: page.data.description ?? '', href: page.url }];
35 });
36
37 -const SITE_DESCRIPTION = `Build AI agents with ${TOOLKIT_COUNT_LABEL} tools. Connect LLMs to external services like GitHub, Slack, Gmail, and more.`;
37 +const SITE_DESCRIPTION = `Build AI agents with ${TOOLKIT_COUNT_LABEL} apps. Connect LLMs to external services like GitHub, Slack, Gmail, and more.`;
38 const HOME_OG_IMAGE = getOgImageUrl('docs', [], 'Composio Docs', HOME_OG_DESCRIPTION);
39
40 export const metadata: Metadata = {
docs/app/llms-full.txt/route.ts
+1 -1
@@ -166,7 +166,7 @@ export async function GET() {
166 ]);
167
168 const results = [
169 - `# Composio Documentation\n\n> Composio powers ${TOOLKIT_COUNT_LABEL} toolkits, tool search, context management, authentication, and a sandboxed workbench to help you build AI agents that turn intent into action.${SESSION_GUARDRAILS}\n\n[Changelog](https://docs.composio.dev/docs/changelog.md): Browse dated release notes and follow each date for its full Markdown.\n# Documentation\n`,
169 + `# Composio Documentation\n\n> Composio powers ${TOOLKIT_COUNT_LABEL} apps, tool search, context management, authentication, and a sandboxed workbench to help you build AI agents that turn intent into action.${SESSION_GUARDRAILS}\n\n[Changelog](https://docs.composio.dev/docs/changelog.md): Browse dated release notes and follow each date for its full Markdown.\n# Documentation\n`,
170 ...docsResults,
171 '\n# Knowledge Hub navigation\n',
172 knowledgeDiscoveryLinks,
docs/app/llms-index.txt/route.ts
+1 -1
@@ -187,7 +187,7 @@ ${legacyReferencePages.map(formatPage).join('\n')}
187
188 const index = `# Composio Documentation
189
190 -> Composio powers ${TOOLKIT_COUNT_LABEL} toolkits, tool search, context management, authentication, and a sandboxed workbench to help you build AI agents that turn intent into action.
190 +> Composio powers ${TOOLKIT_COUNT_LABEL} apps, tool search, context management, authentication, and a sandboxed workbench to help you build AI agents that turn intent into action.
191
192 For a shorter routing map, start at [llms.txt](https://docs.composio.dev/llms.txt).
193
docs/app/llms.txt/route.ts
+1 -1
@@ -12,7 +12,7 @@ export function GET() {
12 .join('\n');
13 return new Response(`# Composio Documentation
14
15 -> Build and operate agents with authentication, tool discovery, execution, and ${TOOLKIT_COUNT_LABEL} toolkits.
15 +> Build and operate agents with authentication, tool discovery, execution, and ${TOOLKIT_COUNT_LABEL} apps.
16
17 Default to Composio Platform for application codebases where users or tenants connect their accounts. Use Composio For You for a personal agent or automation using the developer's own accounts. Start with the current session-based SDK. REST v3.1 is current; REST v3.0 is legacy and is not the starting point for new code.
18
docs/components/home-features.tsx
+1 -1
@@ -59,7 +59,7 @@ export function HomeFeatures() {
59 />
60 <FeatureCard
61 title="Run arbitrary code, safely."
62 - description={`A sandbox pre-wired with your user's connected accounts and ${TOOLKIT_COUNT_LABEL} tools.`}
62 + description={`A sandbox pre-wired with your user's connected accounts and ${TOOLKIT_COUNT_LABEL} apps.`}
63 href="/docs/sandbox"
64 visual={<SandboxVisual />}
65 />
docs/content/docs/claude-code-plugin.mdx
+3 -3
@@ -1,10 +1,10 @@
1 ---
2 title: Claude Code Plugin
3 -description: Act on 1,000+ apps from Claude Code — Slack, GitHub, Gmail, Notion, Linear, and more — with fully managed OAuth. Installs in two commands.
3 +description: Act on %TOOLKIT_COUNT% apps from Claude Code — Slack, GitHub, Gmail, Notion, Linear, and more — with fully managed OAuth. Installs in two commands.
4 keywords: [claude code, plugin, marketplace, composio, oauth, cli, mcp]
5 ---
6
7 -The **Composio plugin for Claude Code** lets Claude act on 1,000+ apps — send the Slack message, open the Linear issue, check your calendar, draft the email. Your agent decides what to do; Composio handles the rest: OAuth, permissions, and finding the right tool for each task. No API keys, no config files.
7 +The **Composio plugin for Claude Code** lets Claude act on %TOOLKIT_COUNT% apps — send the Slack message, open the Linear issue, check your calendar, draft the email. Your agent decides what to do; Composio handles the rest: OAuth, permissions, and finding the right tool for each task. No API keys, no config files.
8
9 <Callout type="info">
10 Using Codex? Follow the [Codex setup](/docs/agent-plugins#configure-one-agent). If you explicitly want MCP in Cursor, Claude Desktop, or another MCP client, use [Composio Connect](/docs/composio-connect).
@@ -79,7 +79,7 @@ Apps connect on demand — the first task that needs one hands you an OAuth link
79 /composio-connect linear
80 ```
81
82 -Works for any of the 1,000+ supported apps — `slack`, `github`, `gmail`, `notion`, `linear`, `hubspot`, and more.
82 +Works for any of the %TOOLKIT_COUNT% supported apps — `slack`, `github`, `gmail`, `notion`, `linear`, `hubspot`, and more.
83
84 ## Team setup
85
docs/content/docs/composio-connect.mdx
+1 -1
@@ -12,7 +12,7 @@ If you are building an application, start with the [SDK Quickstart](/docs/quicks
12
13 ## How Composio Connect works
14
15 -Composio Connect is an MCP server at `https://connect.composio.dev/mcp` that gives your AI agent access to 1000+ apps, including Gmail, Notion, Slack, GitHub, Linear, HubSpot, and Strava, through a single connection.
15 +Composio Connect is an MCP server at `https://connect.composio.dev/mcp` that gives your AI agent access to %TOOLKIT_COUNT% apps, including Gmail, Notion, Slack, GitHub, Linear, HubSpot, and Strava, through a single connection.
16
17 Rather than exposing every app tool directly, Composio exposes **7 meta-tools** that let the agent discover what's available, authorize apps on demand, and execute tools across apps in parallel. The first time your agent needs an app, Composio generates an OAuth link you approve in your browser; after that the connection persists across sessions. See [Available MCP tools](#available-mcp-tools) for the full list.
18
docs/content/docs/migration-guide/mcp-servers-to-sessions.mdx
+1 -1
@@ -22,7 +22,7 @@ Starting fresh? Skip this guide and read [Configuring Sessions](/docs/configurin
22
23 <Callout type="info">
24 Just connecting apps to your own agent for personal use — not building an app? You don't need the
25 -SDK. Use **Composio For You** to connect across 1000+ apps in a few clicks — switch to it from the
25 +SDK. Use **Composio For You** to connect across %TOOLKIT_COUNT% apps in a few clicks — switch to it from the
26 product switcher in the top-left of the dashboard. Reserve `session.mcp.url` for programmatic,
27 in-app use.
28 </Callout>
docs/content/docs/providers/anthropic.mdx
+1 -1
@@ -1,6 +1,6 @@
1 ---
2 title: Anthropic
3 -description: Build an agent with Claude and enable it to use 1000+ tools with Composio.
3 +description: Build an agent with Claude and enable it to use %TOOLKIT_COUNT% apps with Composio.
4 keywords: [claude, messages api, claude agent sdk, agents sdk]
5 ---
6
docs/content/docs/providers/autogen.mdx
+1 -1
@@ -1,6 +1,6 @@
1 ---
2 title: AutoGen
3 -description: Build an agent with AutoGen and enable it to use 1000+ tools with Composio.
3 +description: Build an agent with AutoGen and enable it to use %TOOLKIT_COUNT% apps with Composio.
4 keywords: [autogen, microsoft autogen, ag2]
5 ---
6
docs/content/docs/providers/crewai.mdx
+1 -1
@@ -1,6 +1,6 @@
1 ---
2 title: CrewAI
3 -description: Build an agent with CrewAI and enable it to use 1000+ tools with Composio.
3 +description: Build an agent with CrewAI and enable it to use %TOOLKIT_COUNT% apps with Composio.
4 keywords: [crewai]
5 ---
6
docs/content/docs/providers/custom-providers/index.mdx
+1 -1
@@ -1,6 +1,6 @@
1 ---
2 title: Custom Providers
3 -description: Build an agent with any framework and enable it to use 1000+ tools with Composio.
3 +description: Build an agent with any framework and enable it to use %TOOLKIT_COUNT% apps with Composio.
4 keywords: [custom provider]
5 ---
6
docs/content/docs/providers/custom-providers/python.mdx
+2 -2
@@ -1,11 +1,11 @@
1 ---
2 title: Python Custom Provider
3 -description: Build an agent with any Python framework and enable it to use 1000+ tools with Composio.
3 +description: Build an agent with any Python framework and enable it to use %TOOLKIT_COUNT% apps with Composio.
4 ---
5
6 {/* code-example: single-language-page reason="Python-specific custom provider guide; TypeScript has its own page" */}
7
8 -A **custom provider** adapts Composio tools to the format your AI framework expects, so you can use 1000+ tools with a platform Composio doesn't ship support for. This guide shows you how to build one in Python.
8 +A **custom provider** adapts Composio tools to the format your AI framework expects, so you can use %TOOLKIT_COUNT% toolkits with a platform Composio doesn't ship support for. This guide shows you how to build one in Python.
9
10 ## Provider architecture
11
docs/content/docs/providers/custom-providers/typescript.mdx
+2 -2
@@ -1,11 +1,11 @@
1 ---
2 title: TypeScript Custom Provider
3 -description: Build an agent with any TypeScript framework and enable it to use 1000+ tools with Composio.
3 +description: Build an agent with any TypeScript framework and enable it to use %TOOLKIT_COUNT% apps with Composio.
4 ---
5
6 {/* code-example: single-language-page reason="TypeScript-specific custom provider guide; Python has its own page" */}
7
8 -A **provider** adapts Composio tools to the format your AI framework expects. Write one, and any framework can call Composio's 1000+ tools. This guide shows you how to build your own in TypeScript.
8 +A **provider** adapts Composio tools to the format your AI framework expects. Write one, and any framework can call tools across %TOOLKIT_COUNT% toolkits. This guide shows you how to build your own in TypeScript.
9
10 ## Provider architecture
11
docs/content/docs/providers/google.mdx
+1 -1
@@ -1,6 +1,6 @@
1 ---
2 title: Google
3 -description: Build an agent with Gemini or Google ADK and enable it to use 1000+ tools with Composio.
3 +description: Build an agent with Gemini or Google ADK and enable it to use %TOOLKIT_COUNT% apps with Composio.
4 keywords: [gemini, google generative ai, google genai, google adk, agent development kit]
5 ---
6
docs/content/docs/providers/langchain.mdx
+1 -1
@@ -1,6 +1,6 @@
1 ---
2 title: LangChain
3 -description: Build an agent with LangChain or LangGraph and enable it to use 1000+ tools with Composio.
3 +description: Build an agent with LangChain or LangGraph and enable it to use %TOOLKIT_COUNT% apps with Composio.
4 keywords: [langchain, langgraph]
5 ---
6
docs/content/docs/providers/llamaindex.mdx
+1 -1
@@ -1,6 +1,6 @@
1 ---
2 title: LlamaIndex
3 -description: Build an agent with LlamaIndex and enable it to use 1000+ tools with Composio.
3 +description: Build an agent with LlamaIndex and enable it to use %TOOLKIT_COUNT% apps with Composio.
4 keywords: [llamaindex]
5 ---
6
docs/content/docs/providers/mastra.mdx
+1 -1
@@ -1,6 +1,6 @@
1 ---
2 title: Mastra
3 -description: Build an agent with Mastra and enable it to use 1000+ tools with Composio.
3 +description: Build an agent with Mastra and enable it to use %TOOLKIT_COUNT% apps with Composio.
4 keywords: [mastra]
5 ---
6
docs/content/docs/providers/openai.mdx
+1 -1
@@ -1,6 +1,6 @@
1 ---
2 title: OpenAI
3 -description: Build an agent with OpenAI and enable it to use 1000+ tools with Composio.
3 +description: Build an agent with OpenAI and enable it to use %TOOLKIT_COUNT% apps with Composio.
4 keywords: [gpt, responses api, chat completions, openai agents sdk, agents sdk]
5 ---
6
docs/content/docs/providers/vercel.mdx
+1 -1
@@ -1,6 +1,6 @@
1 ---
2 title: Vercel AI SDK
3 -description: Build an agent with Vercel AI SDK and enable it to use 1000+ tools with Composio.
3 +description: Build an agent with Vercel AI SDK and enable it to use %TOOLKIT_COUNT% apps with Composio.
4 keywords: [ai sdk, vercel]
5 ---
6
docs/content/docs/quickstart.mdx
+1 -1
@@ -9,7 +9,7 @@ Build an agent that chooses Composio tools at runtime. Type a task, connect an a
9 <div className="not-prose my-6 grid grid-cols-1 items-start gap-4 md:grid-cols-2">
10 <div>
11 <div className="mb-6">
12 - Pick your framework below. A Composio [session](/docs/how-composio-works) gives your agent tool discovery, account connections, and execution across [1000+ apps](/toolkits). It exposes only a small set of meta tools, so app schemas are loaded when the agent needs them.
12 + Pick your framework below. A Composio [session](/docs/how-composio-works) gives your agent tool discovery, account connections, and execution across [%TOOLKIT_COUNT% apps](/toolkits). It exposes only a small set of meta tools, so app schemas are loaded when the agent needs them.
13 </div>
14 <AgentSetupActions />
15 </div>
docs/content/examples/general-agent-with-pi.mdx
+1 -1
@@ -10,7 +10,7 @@ gallery:
10 order: 0
11 ---
12
13 -The agent is the easy part. [Pi](https://github.com/earendil-works/pi/tree/main/packages/coding-agent) does the reasoning; Composio gives it 1000+ apps to act on. In three lines you have an agent that can open a PR, check a calendar, or search a Notion workspace for one user.
13 +The agent is the easy part. [Pi](https://github.com/earendil-works/pi/tree/main/packages/coding-agent) does the reasoning; Composio gives it %TOOLKIT_COUNT% apps to act on. In three lines you have an agent that can open a PR, check a calendar, or search a Notion workspace for one user.
14
15 The work is everything around it: putting that agent in Slack, where a whole team talks to it, and making it act as *each* person while posting as one bot. That's a handful of Composio pieces:
16
docs/content/examples/harness-integration.mdx
+1 -1
@@ -10,7 +10,7 @@ gallery:
10 order: 4
11 ---
12
13 -Most Composio examples hand the agent a session and let Composio's [meta tools](/docs/how-composio-works#meta-tools) find, authenticate, and run tools. That default works well until you already have your own planner, retrieval index, dispatcher, and permission layer. Then you need the 1000+ apps and managed auth, not another agent inside yours.
13 +Most Composio examples hand the agent a session and let Composio's [meta tools](/docs/how-composio-works#meta-tools) find, authenticate, and run tools. That default works well until you already have your own planner, retrieval index, dispatcher, and permission layer. Then you need the %TOOLKIT_COUNT% apps and managed auth, not another agent inside yours.
14
15 Composio sits underneath your harness in four calls:
16
docs/content/examples/imessage-agent.mdx
+1 -1
@@ -27,7 +27,7 @@ The pattern is the takeaway, not iMessage specifically. It comes together from a
27
28 1. **A local toolkit** wraps local iMessage (send, contacts, history, memory) as Composio tools.
29 2. **In-process execution** runs each tool right on your Mac through `session.execute`, with no remote API.
30 -3. **One session** puts those local tools on the same surface as the 1000+ app catalog.
30 +3. **One session** puts those local tools on the same surface as the %TOOLKIT_COUNT% toolkit catalog.
31 4. **The eve provider** turns `session.tools()` into [eve](https://github.com/vercel/eve)-native tools, so `eve dev` can call them.
32 5. **Triggers** let an outside event wake the agent and reach your phone, with in-chat auth through `COMPOSIO_MANAGE_CONNECTIONS` for any app you haven't connected.
33
docs/content/examples/local-sandbox-pr-reviewer.mdx
+1 -1
@@ -10,7 +10,7 @@ gallery:
10 order: 2
11 ---
12
13 -Composio usually runs your tools for you. A **local sandbox** is for the times you need to run them yourself: your filesystem, your shell, your security boundary. You still get [managed auth](/docs/authentication) and 1000+ apps; you just keep the code execution.
13 +Composio usually runs your tools for you. A **local sandbox** is for the times you need to run them yourself: your filesystem, your shell, your security boundary. You still get [managed auth](/docs/authentication) and %TOOLKIT_COUNT% apps; you just keep the code execution.
14
15 This example builds a GitHub PR reviewer that does exactly that: it clones a pull request into a sandbox *you* own, runs the repo's real checks there, and posts one grounded comment. The sandbox here is E2B, but E2B is just the sample. The same pattern works with your own VM, container, Kubernetes job, or internal sandbox service.
16
docs/content/kb/guide/toolkits-xero.mdx
+1 -1
@@ -24,7 +24,7 @@ Remove the deprecated/invalid Xero scopes `accounting.journals.read`, `accountin
24
25 ## Connect MCP discovers Xero tools through meta-tools instead of preloading every tool
26
27 -Connect MCP uses meta-tools such as `COMPOSIO_SEARCH_TOOLS` and `COMPOSIO_MULTI_EXECUTE_TOOL` to discover and execute toolkit-specific tools dynamically. For Xero, the expected flow is: ask/search for the task such as `get Xero contacts`, let the agent discover the relevant Xero tool, then execute it through the multi-execute tool. This avoids loading 1000+ tools into context up front.
27 +Connect MCP uses meta-tools such as `COMPOSIO_SEARCH_TOOLS` and `COMPOSIO_MULTI_EXECUTE_TOOL` to discover and execute toolkit-specific tools dynamically. For Xero, the expected flow is: ask/search for the task such as `get Xero contacts`, let the agent discover the relevant Xero tool, then execute it through the multi-execute tool. This avoids loading schemas from every toolkit into context up front.
28
29 ## Connect MCP and Platform MCP Xero connections are independent
30
docs/content/reference/index.mdx
+1 -1
@@ -4,7 +4,7 @@ description: Composio REST API and SDK reference for AI agent tool execution, au
4 keywords: [Composio API, REST API, SDK reference, AI agent tools, tool execution, API documentation]
5 ---
6
7 -Composio powers tool discovery, execution, authentication, and context management for your AI agents with 1000+ toolkits. This reference covers our REST APIs and SDKs.
7 +Composio powers tool discovery, execution, authentication, and context management for your AI agents with %TOOLKIT_COUNT% toolkits. This reference covers our REST APIs and SDKs.
8
9 ## Quick Reference
10
docs/content/reference/v3/index.mdx
+1 -1
@@ -4,7 +4,7 @@ description: Composio REST API and SDK reference for AI agent tool execution, au
4 keywords: [Composio API, REST API, SDK reference, AI agent tools, tool execution, API documentation]
5 ---
6
7 -Composio powers tool discovery, execution, authentication, and context management for your AI agents with 1000+ toolkits. This reference covers our REST APIs and SDKs.
7 +Composio powers tool discovery, execution, authentication, and context management for your AI agents with %TOOLKIT_COUNT% toolkits. This reference covers our REST APIs and SDKs.
8
9 ## Quick Reference
10
docs/content/toolkits/meta-tools/index.mdx
+2 -2
@@ -10,7 +10,7 @@ import { Callout } from 'fumadocs-ui/components/callout';
10
11 Every Composio [session](/docs/how-composio-works) hands your agent a small set of meta tools instead of hundreds of raw tool definitions. The agent uses them to find the right tools for a task, connect the accounts those tools need, execute them, and process the results, all at runtime and all sharing one `session_id`.
12
13 -This keeps your context window small: you load a handful of meta tools, not a catalog of 500+ apps. The agent searches for what it needs when it needs it.
13 +This keeps your context window small: you load a handful of meta tools, not a catalog of %TOOLKIT_COUNT% toolkits. The agent searches for what it needs when it needs it.
14
15 A typical workflow runs in order: call `COMPOSIO_SEARCH_TOOLS` to discover tools and open a session, call `COMPOSIO_MANAGE_CONNECTIONS` if a toolkit is not yet connected, then run the tools with `COMPOSIO_MULTI_EXECUTE_TOOL`. Reach for the workbench and bash tools when responses are large enough to process out of context.
16
@@ -22,7 +22,7 @@ A typical workflow runs in order: call `COMPOSIO_SEARCH_TOOLS` to discover tools
22 | [`COMPOSIO_PROXY_EXECUTE`](/toolkits/meta-tools/proxy_execute) | Sends a raw HTTP request to a connected app's own API, authenticated with your connected account, for endpoints no Composio tool covers |
23 | [`COMPOSIO_REMOTE_BASH_TOOL`](/toolkits/meta-tools/remote_bash_tool) | Runs bash commands in a remote sandbox for file operations, data processing, and system tasks |
24 | [`COMPOSIO_REMOTE_WORKBENCH`](/toolkits/meta-tools/remote_workbench) | Runs Python in a persistent remote sandbox to process large remote files and script bulk or repeated tool executions |
25 -| [`COMPOSIO_SEARCH_TOOLS`](/toolkits/meta-tools/search_tools) | Discovers the right tools across 500+ apps for a task and returns them with an execution plan, connection status, and the `session_id` that ties the rest of the workflow together |
25 +| [`COMPOSIO_SEARCH_TOOLS`](/toolkits/meta-tools/search_tools) | Discovers the right tools across %TOOLKIT_COUNT% toolkits for a task and returns them with an execution plan, connection status, and the `session_id` that ties the rest of the workflow together |
26
27 <Callout type="warn">
28 These schemas are for reference only. We do not guarantee backward compatibility for parameter names or response shapes, so do not rely on them as structured type definitions in your code.
docs/content/toolkits/meta-tools/search_tools.mdx
+1 -1
@@ -1,6 +1,6 @@
1 ---
2 title: Search Tools
3 -description: "Tool Server Info: Composio connects 500+ apps—Slack, GitHub, Notion, Google Workspace (Gmail, Sheets, Drive, Calendar..."
3 +description: "Discovers the right tools across %TOOLKIT_COUNT% toolkits for a task and returns them with an execution plan, connect..."
4 keywords: [COMPOSIO_SEARCH_TOOLS, meta tool]
5 ---
6
docs/kb/articles/toolkits-xero.md
+1 -1
@@ -12,7 +12,7 @@ Remove the deprecated/invalid Xero scopes `accounting.journals.read`, `accountin
12
13 ## Connect MCP discovers Xero tools through meta-tools instead of preloading every tool
14
15 -Connect MCP uses meta-tools such as `COMPOSIO_SEARCH_TOOLS` and `COMPOSIO_MULTI_EXECUTE_TOOL` to discover and execute toolkit-specific tools dynamically. For Xero, the expected flow is: ask/search for the task such as `get Xero contacts`, let the agent discover the relevant Xero tool, then execute it through the multi-execute tool. This avoids loading 1000+ tools into context up front.
15 +Connect MCP uses meta-tools such as `COMPOSIO_SEARCH_TOOLS` and `COMPOSIO_MULTI_EXECUTE_TOOL` to discover and execute toolkit-specific tools dynamically. For Xero, the expected flow is: ask/search for the task such as `get Xero contacts`, let the agent discover the relevant Xero tool, then execute it through the multi-execute tool. This avoids loading schemas from every toolkit into context up front.
16
17 ## Connect MCP and Platform MCP Xero connections are independent
18
docs/lib/home-navigation.ts
+1
@@ -201,6 +201,7 @@ export const DOCS_PRODUCTS = {
201 url: '/docs/authentication/managing-multiple-connected-accounts',
202 label: 'Multiple connected accounts',
203 },
204 + { url: '/docs/authentication/session-cant-find-account' },
205 { url: '/docs/authentication/controlling-scopes' },
206 { url: '/docs/authentication/manually-authenticating' },
207 { url: '/docs/authentication/programmatic-auth-configs' },
docs/lib/meta-tool-overrides.ts
+5 -2
@@ -1,3 +1,5 @@
1 +import { resolveToolkitCount } from './toolkit-count';
2 +
3 /**
4 * Hand-written copy overrides for meta tool reference pages.
5 *
@@ -30,7 +32,7 @@ export const META_TOOL_OVERRIDES: Record<string, MetaToolOverride> = {
32 },
33 COMPOSIO_SEARCH_TOOLS: {
34 summary:
33 - 'Discovers the right tools across 500+ apps for a task and returns them with an execution plan, connection status, and the `session_id` that ties the rest of the workflow together.',
35 + 'Discovers the right tools across %TOOLKIT_COUNT% toolkits for a task and returns them with an execution plan, connection status, and the `session_id` that ties the rest of the workflow together.',
36 whenToUse:
37 'Call `COMPOSIO_SEARCH_TOOLS` first, at the start of any task that touches an external app or service. Run it again whenever the user pivots to a new use case so you get a fresh `session_id` and a plan scoped to the new work.',
38 usageNote:
@@ -79,5 +81,6 @@ export const META_TOOL_OVERRIDES: Record<string, MetaToolOverride> = {
81 };
82
83 export function getMetaToolOverride(slug: string): MetaToolOverride | undefined {
82 - return META_TOOL_OVERRIDES[slug];
84 + const override = META_TOOL_OVERRIDES[slug];
85 + return override ? { ...override, summary: resolveToolkitCount(override.summary) } : undefined;
86 }
docs/lib/search-index.ts
+2 -1
@@ -1,4 +1,5 @@
1 // Use direct imports from collections to avoid top-level await in lib/source.ts.
2 +import { resolveToolkitCount } from './toolkit-count';
3 import { createHash } from 'node:crypto';
4 import { existsSync, readdirSync, readFileSync } from 'node:fs';
5 import { join, relative } from 'node:path';
@@ -541,7 +542,7 @@ function getFilesystemRecords(): AlgoliaDocsRecord[] {
542 if (!route) return [];
543 if (isExcludedFromSearch(route.url)) return [];
544
544 - const source = readFileSync(file, 'utf8');
545 + const source = resolveToolkitCount(readFileSync(file, 'utf8'));
546 const frontmatter = getFrontmatter(source);
547 const title = getFrontmatterValue(frontmatter, 'title');
548 if (!title) return [];
docs/lib/source.ts
+2 -2
@@ -30,7 +30,7 @@ import { PACKAGE_MANAGERS } from './package-install';
30 import { z } from 'zod';
31 import { promptFor, SETUP_PROMPT } from './agent-prompts';
32 import { AGENTS } from './agent-setup-clients';
33 -import { HOME_OG_DESCRIPTION } from './toolkit-count';
33 +import { HOME_OG_DESCRIPTION, resolveToolkitCount } from './toolkit-count';
34 import { componentsToMarkdown } from './example-document';
35 import { providerListMarkdown } from './providers';
36
@@ -355,7 +355,7 @@ function endpointsTableToMarkdown(payload: string, version: ApiVersion, url?: st
355 * there since the changelog tree is not versioned.
356 */
357 export function mdxToCleanMarkdown(content: string, url?: string): string {
358 - let result = content;
358 + let result = resolveToolkitCount(content);
359 const version = url ? detectApiVersion(url) : '3.1';
360
361 // Remove frontmatter
docs/lib/toolkit-count.ts
+5 -13
@@ -1,16 +1,8 @@
1 -import toolkitsData from '@/public/data/toolkits-list.json';
1 +// Update the docs catalog claim here. Markdown and MDX use %TOOLKIT_COUNT%.
2 +export const TOOLKIT_COUNT_LABEL = '1,500+';
3
3 -// Server-only: reads the full toolkit catalog at build time to derive a
4 -// rounded-down, human-friendly count for marketing copy (hero, metadata,
5 -// llms.txt). Never import this from a "use client" module — that would ship
6 -// the ~1,300-row JSON array to the browser just to read its length.
7 -const roundedToolkitCount = Math.floor(toolkitsData.length / 100) * 100;
4 +export function resolveToolkitCount(text: string): string {
5 + return text.replaceAll('%TOOLKIT_COUNT%', TOOLKIT_COUNT_LABEL);
6 +}
7
9 -// Pin the locale so the grouping separator is a comma regardless of the build
10 -// host's locale ("1,300+", never "1.300+" or "1 300+").
11 -export const TOOLKIT_COUNT_LABEL = `${roundedToolkitCount.toLocaleString('en-US')}+`;
12 -
13 -// Copy for the docs homepage social card. Counts apps rather than tools so
14 -// it matches the landing page. Every route that builds the home-card URL
15 -// uses this, so the live count survives whichever page sets the metadata.
8 export const HOME_OG_DESCRIPTION = `Build AI agents with ${TOOLKIT_COUNT_LABEL} apps. Connect LLMs to external services like GitHub, Slack, Gmail, and more.`;
docs/package.json
+1 -1
@@ -41,7 +41,7 @@
41 "check:python-snippets": "bun scripts/check-python-snippets.ts",
42 "test:python-snippets": "bun scripts/python-snippets/test-helper.ts",
43 "check:typescript-snippets": "bun scripts/check-typescript-snippets.ts",
44 - "pretest": "bun run generate:kb",
44 + "pretest": "bun run generate:kb && bun run build:agent-index",
45 "test": "bun test tests/static/",
46 "test:integration": "bun test tests/integration/ --timeout 30000"
47 },
docs/scripts/build-agent-index.ts
+2 -1
@@ -19,6 +19,7 @@
19 *
20 * Run manually with `bun scripts/build-agent-index.ts`.
21 */
22 +import { resolveToolkitCount } from '../lib/toolkit-count';
23 import { existsSync, readdirSync, readFileSync, writeFileSync } from 'node:fs';
24 import { join, relative } from 'node:path';
25
@@ -306,7 +307,7 @@ for (const collection of COLLECTIONS) {
307 const absPath = join(dir, entry);
308 const url = urlFromContentPath(absPath);
309 if (!url) continue;
309 - const raw = readFileSync(absPath, 'utf8');
310 + const raw = resolveToolkitCount(readFileSync(absPath, 'utf8'));
311 const { title, description, legacy } = parseFrontmatter(raw);
312 pages.push({
313 collection,
docs/scripts/generate-meta-tools.ts
+6 -2
@@ -199,7 +199,7 @@ import { Callout } from 'fumadocs-ui/components/callout';
199
200 Every Composio [session](/docs/how-composio-works) hands your agent a small set of meta tools instead of hundreds of raw tool definitions. The agent uses them to find the right tools for a task, connect the accounts those tools need, execute them, and process the results, all at runtime and all sharing one \`session_id\`.
201
202 -This keeps your context window small: you load a handful of meta tools, not a catalog of 500+ apps. The agent searches for what it needs when it needs it.
202 +This keeps your context window small: you load a handful of meta tools, not a catalog of %TOOLKIT_COUNT% toolkits. The agent searches for what it needs when it needs it.
203
204 A typical workflow runs in order: call \`COMPOSIO_SEARCH_TOOLS\` to discover tools and open a session, call \`COMPOSIO_MANAGE_CONNECTIONS\` if a toolkit is not yet connected, then run the tools with \`COMPOSIO_MULTI_EXECUTE_TOOL\`. Reach for the workbench and bash tools when responses are large enough to process out of context.
205
@@ -222,7 +222,11 @@ These schemas are for reference only. We do not guarantee backward compatibility
222
223 /** Generate an individual tool MDX page */
224 export function generateToolMdx(tool: GeneratedMetaTool): string {
225 - const desc = encodeYamlString(briefDescription(tool.description));
225 + // The API's search-tool description contains an outdated catalog count.
226 + const description = tool.slug === 'COMPOSIO_SEARCH_TOOLS'
227 + ? META_TOOL_OVERRIDES[tool.slug].summary
228 + : tool.description;
229 + const desc = encodeYamlString(briefDescription(description));
230
231 return `---
232 title: ${tool.displayName}
docs/source.config.ts
+13
@@ -10,12 +10,15 @@ import { transformerTwoslash } from '@shikijs/twoslash';
10 import { createFileSystemTypesCache } from '@shikijs/vitepress-twoslash/cache-fs';
11 import { remarkMdxMermaid } from 'fumadocs-core/mdx-plugins';
12 import { z } from 'zod';
13 +import { resolveToolkitCount } from './lib/toolkit-count';
14
15 // You can customise Zod schemas for frontmatter and `meta.json` here
16 // see https://fumadocs.dev/docs/mdx/collections
17
18 // Extended schema with keywords for search
19 const docsSchema = frontmatterSchema.extend({
20 + title: z.string().transform(resolveToolkitCount),
21 + description: z.string().transform(resolveToolkitCount).optional(),
22 keywords: z.array(z.string()).optional(),
23 productAreas: z
24 .array(
@@ -202,6 +205,16 @@ export const changelog = defineCollections({
205 });
206
207 export default defineConfig({
208 + plugins: [
209 + {
210 + name: 'toolkit-count',
211 + doc: {
212 + vfile(file) {
213 + file.value = resolveToolkitCount(String(file.value));
214 + },
215 + },
216 + },
217 + ],
218 mdxOptions: {
219 remarkPlugins: [remarkMdxMermaid],
220 rehypeCodeOptions: {
docs/tests/integration/toolkit-count.test.ts new
+55
@@ -0,0 +1,55 @@
1 +import { expect, test } from 'bun:test';
2 +import { TOOLKIT_COUNT_LABEL } from '../../lib/toolkit-count';
3 +import { fetchPage } from './helpers';
4 +
5 +async function paragraphText(html: string): Promise<string> {
6 + let content = '';
7 + await new HTMLRewriter()
8 + .on('main p', {
9 + text(chunk) {
10 + content += chunk.text;
11 + },
12 + })
13 + .transform(new Response(html))
14 + .text();
15 + return content;
16 +}
17 +
18 +test('paragraph extraction ignores metadata and script contents', async () => {
19 + for (const closeTag of ['</script>', '</script >']) {
20 + const html = `<head><meta name="description" content="1,500+ apps"></head>
21 + <body><main><script>const text = '<p>1,500+ apps</p>';${closeTag}
22 + <p>1,500+ <strong>toolkits</strong></p></main></body>`;
23 + expect(await paragraphText(html)).toBe('1,500+ toolkits');
24 + }
25 +});
26 +
27 +test('catalog claims use apps in marketing copy and toolkits in technical docs', async () => {
28 + for (const [path, noun] of [
29 + ['/docs', 'apps'],
30 + ['/docs/quickstart', 'apps'],
31 + ['/docs/providers/openai', 'apps'],
32 + ['/docs/claude-code-plugin', 'apps'],
33 + ['/reference', 'toolkits'],
34 + ['/toolkits/meta-tools', 'toolkits'],
35 + ['/toolkits/meta-tools/search_tools', 'toolkits'],
36 + ['/docs/quickstart.md', 'apps'],
37 + ['/reference.md', 'toolkits'],
38 + ['/llms.txt', 'apps'],
39 + ]) {
40 + const response = await fetchPage(path);
41 + expect(response.status, path).toBe(200);
42 + const text = await response.text();
43 + // Parse authored paragraphs so metadata and hydration payloads cannot satisfy body assertions.
44 + const body = response.headers.get('content-type')?.includes('text/html')
45 + ? await paragraphText(text)
46 + : text;
47 + if (path !== '/docs/providers/openai') {
48 + expect(body, path).toContain(`${TOOLKIT_COUNT_LABEL} ${noun}`);
49 + }
50 + expect(text, path).not.toContain('%TOOLKIT_COUNT%');
51 + if (path === '/docs/providers/openai') {
52 + expect(text).toContain(`name="description" content="Build an agent with OpenAI and enable it to use ${TOOLKIT_COUNT_LABEL} apps with Composio.`);
53 + }
54 + }
55 +});
docs/tests/static/product-navigation.test.ts
+1
@@ -131,6 +131,7 @@ describe('Docs product navigation', () => {
131 '/docs/providers',
132 '/docs/how-composio-works',
133 '/docs/authentication',
134 + '/docs/authentication/session-cant-find-account',
135 '/docs/skills',
136 '/docs/triggers',
137 '/docs/consumer-agents',
docs/tests/static/toolkit-count.test.ts new
+36
@@ -0,0 +1,36 @@
1 +import { expect, test } from 'bun:test';
2 +import { readFileSync } from 'node:fs';
3 +import { frontmatter } from 'fumadocs-core/content/md/frontmatter';
4 +import { docs } from '../../source.config';
5 +import { readPageByUrl } from '../../agent/lib/docs';
6 +import context from '../../agent/instructions/context';
7 +import { TOOLKIT_COUNT_LABEL, resolveToolkitCount } from '../../lib/toolkit-count';
8 +import { getMetaToolBySlug } from '../../lib/meta-tools-data';
9 +import { generateIndexMdx, generateToolMdx } from '../../scripts/generate-meta-tools';
10 +
11 +test('one catalog count reaches frontmatter, agent content, and meta-tool display copy', async () => {
12 + expect(resolveToolkitCount('%TOOLKIT_COUNT% toolkits, %TOOLKIT_COUNT% apps'))
13 + .toBe(`${TOOLKIT_COUNT_LABEL} toolkits, ${TOOLKIT_COUNT_LABEL} apps`);
14 + for (const [path, noun] of [
15 + ['content/docs/providers/openai.mdx', 'apps'],
16 + ['content/docs/claude-code-plugin.mdx', 'apps'],
17 + ]) {
18 + const parsed = docs.docs.schema.parse(frontmatter(readFileSync(path, 'utf8')).data);
19 + expect(parsed.description).toContain(`${TOOLKIT_COUNT_LABEL} ${noun}`);
20 + }
21 + expect(readPageByUrl('/docs/quickstart')?.raw).toContain(`${TOOLKIT_COUNT_LABEL} apps`);
22 + expect(readPageByUrl('/reference')?.raw).toContain(`${TOOLKIT_COUNT_LABEL} toolkits`);
23 + expect(readPageByUrl('/docs/providers/custom-providers/typescript')?.raw)
24 + .toContain(`call tools across ${TOOLKIT_COUNT_LABEL} toolkits`);
25 + expect(context.markdown).toContain(`${TOOLKIT_COUNT_LABEL} toolkits`);
26 + expect((await getMetaToolBySlug('COMPOSIO_SEARCH_TOOLS'))?.summary)
27 + .toContain(`${TOOLKIT_COUNT_LABEL} toolkits`);
28 +});
29 +
30 +test('meta-tool regeneration retains the count placeholder instead of the API count', () => {
31 + const tools = JSON.parse(readFileSync('public/data/meta-tools.json', 'utf8'));
32 + expect(generateIndexMdx(tools)).toBe(readFileSync('content/toolkits/meta-tools/index.mdx', 'utf8'));
33 + const search = tools.find((tool: { slug: string }) => tool.slug === 'COMPOSIO_SEARCH_TOOLS');
34 + expect(generateToolMdx(search)).toBe(readFileSync('content/toolkits/meta-tools/search_tools.mdx', 'utf8'));
35 + expect(generateToolMdx(search)).toContain('%TOOLKIT_COUNT% toolkits');
36 +});