Composio powers 1000+ toolkits, tool search, context management, authentication, and a sandboxed workbench to help you build AI agents that turn intent into action.
docs: Rewrite Vercel AI SDK provider page — Steps pattern, tested e2e
- Standardize to Steps pattern (Install → Configure → Run) matching other providers - Use IntegrationTabs for Tools vs MCP sections - Remove dotenv imports and verbose console.log for consistency - Add note about Vercel being TS-only and agentic (no manual loop) - Update Google model to gemini-3-pro-preview - Both Vercel examples (Tools + MCP) tested end-to-end Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Sushmithamallesh committed
Feb 17, 2026 at 14:42 UTC
57c9ccb14ec03f2257bf93f131c22b0ebd6ee59d
2 files changed
+66
-68
docs/content/docs/providers/google.mdx
+2
-2
@@ -56,7 +56,7 @@ session = composio.create(user_id="user_123")
56
tools = session.tools()
57
58
config = types.GenerateContentConfig(tools=tools)
59
-chat = client.chats.create(model="gemini-2.0-flash", config=config)
59
+chat = client.chats.create(model="gemini-3-pro-preview", config=config)
60
response = chat.send_message("Star the composiohq/composio repository on github")
61
62
# Agentic loop — keep executing tool calls until the model responds with text
@@ -87,7 +87,7 @@ const session = await composio.create("user_123");
87
const tools = await session.tools();
88
89
const chat = ai.chats.create({
90
- model: 'gemini-2.0-flash',
90
+ model: 'gemini-3-pro-preview',
91
config: {
92
tools: [{ functionDeclarations: tools }],
93
},
docs/content/docs/providers/vercel.mdx
+64
-66
@@ -1,140 +1,138 @@
1
---
2
title: Vercel AI SDK
3
description: Build an agent with Vercel AI SDK and enable it to use 800+ tools with Composio.
4
-keywords: [ai sdk]
4
+keywords: [ai sdk, vercel]
5
---
6
7
-Vercel AI SDK allows you to configure an optional async `execute` function that the framework uses to execute the tool calls.
7
+The Vercel AI SDK provider transforms Composio tools into Vercel's [tool format](https://sdk.vercel.ai/docs/ai-sdk-core/tools-and-tool-calling) with built-in execution — no manual agentic loop needed.
8
9
-The Vercel provider for Composio formats the Composio tools and adds this `execute` function to the tool calls.
10
-
11
-## Usage with Tools
9
+<Callout type="info">
10
+Vercel AI SDK is TypeScript only. There is no Python SDK.
11
+</Callout>
12
13
-Use Composio as a native tool with the Vercel AI SDK.
13
+<IntegrationTabs tabs={[
14
+ { value: "tools", label: "Tools" },
15
+ { value: "mcp", label: "MCP" },
16
+]}>
17
15
-### Installation
18
+<IntegrationContent value="tools">
19
+<Steps>
20
+<Step>
21
+**Install**
22
23
```bash
24
npm install @composio/core @composio/vercel ai @ai-sdk/anthropic
25
```
26
+</Step>
27
21
-### Usage
22
-
23
-Create a session and use it as a native tool with Vercel AI SDK:
28
+<Step>
29
+**Configure API Keys**
30
31
<Callout type="info">
26
-- Set `COMPOSIO_API_KEY` environment variable with your API key from [Settings](https://platform.composio.dev/?next_page=/settings).
27
-- Set `ANTHROPIC_API_KEY` environment variable with your [Anthropic API key](https://console.anthropic.com/settings/keys).
32
+Set `COMPOSIO_API_KEY` with your API key from [Settings](https://platform.composio.dev/?next_page=/settings) and `ANTHROPIC_API_KEY` with your [Anthropic API key](https://console.anthropic.com/settings/keys).
33
</Callout>
34
35
+```txt title=".env"
36
+COMPOSIO_API_KEY=xxxxxxxxx
37
+ANTHROPIC_API_KEY=xxxxxxxxx
38
+```
39
+</Step>
40
+
41
+<Step>
42
+**Create session and run**
43
+
44
+The Vercel provider is **agentic** — tools include an `execute` function, so the AI SDK handles tool calls automatically via `maxSteps`.
45
+
46
```typescript
47
// @noErrors
32
-import "dotenv/config";
48
import { anthropic } from "@ai-sdk/anthropic";
49
import { Composio } from "@composio/core";
50
import { VercelProvider } from "@composio/vercel";
36
-import { stepCountIs, streamText } from "ai";
51
+import { streamText, stepCountIs } from "ai";
52
38
-// Initialize Composio with Vercel provider (API key from env var COMPOSIO_API_KEY)
53
const composio = new Composio({ provider: new VercelProvider() });
54
41
-// Unique identifier of the user
42
-const userId = "user_123";
43
-
44
-// Create a session and get native tools for the user
45
-const session = await composio.create(userId);
55
+// Create a session for your user
56
+const session = await composio.create("user_123");
57
const tools = await session.tools();
58
48
-console.log("Fetching GitHub issues from the Composio repository...");
49
-
50
-// Stream the response with tool calling
59
const stream = await streamText({
52
- system: "You are a helpful personal assistant. Use Composio tools to take action.",
60
model: anthropic("claude-sonnet-4-5"),
54
- prompt: "Fetch all the open GitHub issues on the composio repository and group them by bugs/features/docs.",
55
- stopWhen: stepCountIs(10),
56
- onStepFinish: (step) => {
57
- for (const toolCall of step.toolCalls) {
58
- console.log(`[Using tool: ${toolCall.toolName}]`);
59
- }
60
- },
61
tools,
62
+ prompt: "Star the composiohq/composio repository on github",
63
+ stopWhen: stepCountIs(10),
64
});
65
66
for await (const textPart of stream.textStream) {
67
process.stdout.write(textPart);
68
}
67
-
68
-console.log("\n\n---");
69
-console.log("Tip: If prompted to authenticate, complete the auth flow and run again.");
69
```
70
+</Step>
71
+</Steps>
72
+</IntegrationContent>
73
72
-## Usage with MCP
73
-
74
-Use Composio with the Vercel AI SDK for a fully managed MCP experience.
74
+<IntegrationContent value="mcp">
75
+<Steps>
76
+<Step>
77
+**Install**
78
76
-### Installation
79
+No `@composio/vercel` package needed — use Vercel AI SDK's [MCP client](https://sdk.vercel.ai/docs/ai-sdk-core/mcp) directly.
80
81
```bash
79
-npm install dotenv @composio/core ai @ai-sdk/anthropic @ai-sdk/mcp
82
+npm install @composio/core ai @ai-sdk/anthropic @ai-sdk/mcp
83
```
84
+</Step>
85
82
-### Usage
83
-
84
-Use Composio with Vercel AI SDK's `streamText` for completions:
86
+<Step>
87
+**Configure API Keys**
88
89
<Callout type="info">
87
-- Set `COMPOSIO_API_KEY` environment variable with your API key from [Settings](https://platform.composio.dev/?next_page=/settings).
88
-- Set `ANTHROPIC_API_KEY` environment variable with your [Anthropic API key](https://platform.claude.com/settings/keys).
90
+Set `COMPOSIO_API_KEY` with your API key from [Settings](https://platform.composio.dev/?next_page=/settings) and `ANTHROPIC_API_KEY` with your [Anthropic API key](https://console.anthropic.com/settings/keys).
91
</Callout>
92
93
+```txt title=".env"
94
+COMPOSIO_API_KEY=xxxxxxxxx
95
+ANTHROPIC_API_KEY=xxxxxxxxx
96
+```
97
+</Step>
98
+
99
+<Step>
100
+**Create session and run**
101
+
102
```typescript
103
// @noErrors
93
-import "dotenv/config";
104
import { anthropic } from "@ai-sdk/anthropic";
105
import { experimental_createMCPClient as createMCPClient } from "@ai-sdk/mcp";
106
import { Composio } from "@composio/core";
97
-import { stepCountIs, streamText } from "ai";
107
+import { streamText, stepCountIs } from "ai";
108
99
-// Initialize Composio (API key from env var COMPOSIO_API_KEY or pass explicitly: { apiKey: "your-key" })
109
const composio = new Composio();
110
102
-// Unique identifier of the user
103
-const userId = "user_123";
111
+// Create a session for your user
112
+const { mcp } = await composio.create("user_123");
113
105
-// Create a session for the user
106
-const { mcp } = await composio.create(userId);
107
-
108
-// Create an MCP client to connect to the Composio MCP server
114
const client = await createMCPClient({
115
transport: {
116
type: "http",
117
url: mcp.url,
113
- headers: mcp.headers, // Authentication headers for the Composio MCP server
118
+ headers: mcp.headers,
119
},
120
});
121
const tools = await client.tools();
122
118
-console.log("Summarizing your emails from today");
119
-
123
const stream = await streamText({
121
- system: "You are a helpful personal assistant. Use Composio tools to take action.",
124
model: anthropic("claude-sonnet-4-5"),
125
+ tools,
126
prompt: "Summarize my emails from today",
127
stopWhen: stepCountIs(10),
125
- onStepFinish: (step) => {
126
- for (const toolCall of step.toolCalls) {
127
- console.log(`[Using tool: ${toolCall.toolName}]`);
128
- }
129
- },
130
- tools,
128
});
129
130
for await (const textPart of stream.textStream) {
131
process.stdout.write(textPart);
132
}
136
-
137
-console.log("\n\n---");
138
-console.log("Tip: If prompted to authenticate, complete the auth flow and run again.")
133
```
134
+</Step>
135
+</Steps>
136
+</IntegrationContent>
137
138
+</IntegrationTabs>