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: rename the approvals example and move it to the end
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Mukund committed
Oct 15, 2026 at 18:10 UTC
0b5419912d81f4ac48d8e5a1511ab108739fcf8e
1 file changed
+46
-46
docs/content/docs/approvals.mdx
+46
-46
@@ -195,7 +195,52 @@ const result = await session.execute('GMAIL_SEND_EMAIL', args, {
195
196
When the user picks **Always approve**, their approval lasts for the rest of the session. Later calls to that tool on the same connected account run without asking again.
197
198
-## End-to-end example
198
+## Approvals in MCP clients
199
+
200
+Approvals use MCP [elicitation](https://modelcontextprotocol.io/specification/2026-07-28/client/elicitation), which lets a server ask the user for input through the client. When your agent connects to the session [over MCP](/docs/sessions-via-mcp), clients that support elicitation, like Claude Code, ask the user for you. With clients that don't, calls that need approval are denied. To run them without approval instead, set `no_elicitation_support_fallback` to `"allow"`:
201
+
202
+<Tabs groupId="language" items={['Python', 'TypeScript']} persist>
203
+<Tab value="Python">
204
+```python
205
+session = composio.create(
206
+ user_id="user_123",
207
+ tools={
208
+ "gmail": {
209
+ "enable": ["GMAIL_FETCH_EMAILS", "GMAIL_SEND_EMAIL"],
210
+ "require_approval": ["GMAIL_SEND_EMAIL"],
211
+ }
212
+ },
213
+ experimental={"no_elicitation_support_fallback": "allow"},
214
+)
215
+```
216
+</Tab>
217
+<Tab value="TypeScript">
218
+```typescript
219
+// @noErrors
220
+import { Composio } from '@composio/core';
221
+const composio = new Composio({ apiKey: 'your_api_key' });
222
+// ---cut---
223
+const session = await composio.create('user_123', {
224
+ tools: {
225
+ gmail: {
226
+ enable: ['GMAIL_FETCH_EMAILS', 'GMAIL_SEND_EMAIL'],
227
+ requireApproval: ['GMAIL_SEND_EMAIL'],
228
+ },
229
+ },
230
+ experimental: { noElicitationSupportFallback: 'allow' },
231
+});
232
+```
233
+</Tab>
234
+</Tabs>
235
+
236
+## Limitations
237
+
238
+- **Workbench and bash scripts.** Composio reviews `COMPOSIO_REMOTE_WORKBENCH` and `COMPOSIO_REMOTE_BASH_TOOL` scripts on a best-effort basis to find the tools they call. If any of those tools need approval, the script picks tools while it runs, or the review fails, the user approves the whole script. Calls the script makes aren't asked about again, and there's no **Always approve**.
239
+- **Proxy execute.** A proxy request can't be matched to a tool, so `session.proxy_execute()` is disabled for a toolkit when any of its tools can need approval.
240
+- **Tags.** Rules based on [tags](/docs/tool-tags) are only as accurate as the tags, which Composio assigns on a best-effort basis. List tools explicitly for anything you must gate.
241
+- **Providers.** `handle_tool_calls` and framework providers don't handle approval yet. Run tool calls with `session.execute()`, as in the [full example](#full-example) below.
242
+
243
+## Full example
244
245
An agent loop with Claude that asks for approval in the terminal:
246
@@ -380,51 +425,6 @@ terminal.close();
425
</Tab>
426
</Tabs>
427
383
-## Approvals in MCP clients
384
-
385
-Approvals use MCP [elicitation](https://modelcontextprotocol.io/specification/2026-07-28/client/elicitation), which lets a server ask the user for input through the client. When your agent connects to the session [over MCP](/docs/sessions-via-mcp), clients that support elicitation, like Claude Code, ask the user for you. With clients that don't, calls that need approval are denied. To run them without approval instead, set `no_elicitation_support_fallback` to `"allow"`:
386
-
387
-<Tabs groupId="language" items={['Python', 'TypeScript']} persist>
388
-<Tab value="Python">
389
-```python
390
-session = composio.create(
391
- user_id="user_123",
392
- tools={
393
- "gmail": {
394
- "enable": ["GMAIL_FETCH_EMAILS", "GMAIL_SEND_EMAIL"],
395
- "require_approval": ["GMAIL_SEND_EMAIL"],
396
- }
397
- },
398
- experimental={"no_elicitation_support_fallback": "allow"},
399
-)
400
-```
401
-</Tab>
402
-<Tab value="TypeScript">
403
-```typescript
404
-// @noErrors
405
-import { Composio } from '@composio/core';
406
-const composio = new Composio({ apiKey: 'your_api_key' });
407
-// ---cut---
408
-const session = await composio.create('user_123', {
409
- tools: {
410
- gmail: {
411
- enable: ['GMAIL_FETCH_EMAILS', 'GMAIL_SEND_EMAIL'],
412
- requireApproval: ['GMAIL_SEND_EMAIL'],
413
- },
414
- },
415
- experimental: { noElicitationSupportFallback: 'allow' },
416
-});
417
-```
418
-</Tab>
419
-</Tabs>
420
-
421
-## Limitations
422
-
423
-- **Workbench and bash scripts.** Composio reviews `COMPOSIO_REMOTE_WORKBENCH` and `COMPOSIO_REMOTE_BASH_TOOL` scripts on a best-effort basis to find the tools they call. If any of those tools need approval, the script picks tools while it runs, or the review fails, the user approves the whole script. Calls the script makes aren't asked about again, and there's no **Always approve**.
424
-- **Proxy execute.** A proxy request can't be matched to a tool, so `session.proxy_execute()` is disabled for a toolkit when any of its tools can need approval.
425
-- **Tags.** Rules based on [tags](/docs/tool-tags) are only as accurate as the tags, which Composio assigns on a best-effort basis. List tools explicitly for anything you must gate.
426
-- **Providers.** `handle_tool_calls` and framework providers don't handle approval yet. Run tool calls with `session.execute()`, as in the example above.
427
-
428
## Next
429
430
<Card