Composio powers 1000+ toolkits, tool search, context management, authentication, and a sandboxed workbench to help you build AI agents that turn intent into action.
update
Sushmithamallesh committed
Dec 10, 2025 at 16:55 UTC
a2029ebfe5d5fb590a32233009daa390ec35ac14
4 files changed
+56
-17
fern/docs.yml
-9
@@ -320,18 +320,9 @@ navigation:
320
- page: Configuring Custom Auth Configs
321
path: pages/dist/tool-router/configuring-custom-auth-configs.mdx
322
slug: configuring-custom-auth-configs
323
- - page: Configuring Workbench
324
- path: pages/dist/tool-router/configuring-workbench.mdx
325
- slug: configuring-workbench
323
- page: Using with MCP Clients
324
path: pages/dist/tool-router/using-with-mcp-clients.mdx
325
slug: using-with-mcp-clients
329
- # - page: Using as Native Tools
330
- # path: pages/dist/tool-router/using-as-native-tools.mdx
331
- # slug: using-as-native-tools
332
- - page: Handling Files
333
- path: pages/dist/tool-router/file-upload-download.mdx
334
- slug: file-upload-download
326
- tab: providers
327
layout:
328
- section: Supported Providers
fern/pages/src/tool-router/quickstart.mdx
+15
-6
@@ -6,16 +6,21 @@ keywords: "tool router quickstart, quickstart"
6
hide-nav-links: false
7
---
8
9
-Build an agent that summarizes your Gmail inbox.
9
+Build an interactive agent with access to 1000+ tools.
10
11
<Tabs>
12
-<Tab title="OpenAI (TypeScript)">
12
+<Tab title="OpenAI Agents SDK">
13
14
<Steps>
15
-<Step title="Install Composio SDK">
16
-```bash
15
+<Step title="Install">
16
+<CodeGroup>
17
+```bash title="Python" for="python"
18
+pip install composio openai-agents
19
+```
20
+```bash title="TypeScript" for="typescript"
21
npm install @composio/core @openai/agents
22
```
23
+</CodeGroup>
24
</Step>
25
26
<Step title="Create session and run agent">
@@ -29,13 +34,17 @@ npm install @composio/core @openai/agents
34
35
When the agent needs authentication, it will prompt the user with a link to connect their account. If you want to customise the authentication flow and have more control over the user experience, see [Manual Authentication](/tool-router/manually-authenticating-users).
36
32
-<SnippetCode src="fern/snippets/tool-router/typescript/quickstart-openai.ts" />
37
+<CodeGroup>
38
+<SnippetCode src="fern/snippets/tool-router/python/quickstart-openai.py" title="Python" for="python" />
39
+<SnippetCode src="fern/snippets/tool-router/typescript/quickstart-openai.ts" title="TypeScript" for="typescript" />
40
+</CodeGroup>
41
+
42
</Step>
43
</Steps>
44
45
</Tab>
46
38
-<Tab title="Vercel AI SDK (Anthropic)">
47
+<Tab title="Vercel AI SDK">
48
49
<Steps>
50
<Step title="Install Composio SDK">
fern/snippets/tool-router/python/quickstart-openai.py
new
+38
@@ -0,0 +1,38 @@
1
+import os
2
+from composio import Composio
3
+from agents import Agent, Runner, HostedMCPTool, ModelSettings
4
+
5
+composio_api_key = os.environ.get("COMPOSIO_API_KEY")
6
+user_id = "user_123" # Your user's unique identifier
7
+
8
+composio = Composio()
9
+session = composio.create(user_id=user_id)
10
+
11
+agent = Agent(
12
+ name="Personal Assistant",
13
+ instructions="You are a helpful personal assistant. Use Composio tools to take action.",
14
+ model="gpt-5.1",
15
+ model_settings= ModelSettings(
16
+ reasoning={"effort": "low"},
17
+ ),
18
+ tools=[
19
+ HostedMCPTool(
20
+ tool_config={
21
+ "type": "mcp",
22
+ "server_label": "composio",
23
+ "server_url": session.mcp.url,
24
+ "require_approval": "never",
25
+ "headers": {"x-api-key": composio_api_key},
26
+ }
27
+ )
28
+ ],
29
+)
30
+
31
+print("Assistant: What would you like me to do today?\n")
32
+
33
+while True:
34
+ user_input = input("> ")
35
+ if user_input == "exit":
36
+ break
37
+ result = Runner.run_sync(starting_agent=agent, input=user_input)
38
+ print(f"Assistant: {result.final_output}\n")
fern/snippets/tool-router/typescript/quickstart-openai.ts
+3
-2
@@ -27,10 +27,11 @@ const agent = new Agent({
27
});
28
29
const rl = createInterface({ input: process.stdin, output: process.stdout });
30
+console.log("Assistant: What would you like me to do today?\n");
31
while (true) {
31
- const input = await rl.question("You: ");
32
+ const input = await rl.question("> ");
33
if (input === "exit") break;
34
const result = await run(agent, input);
34
- console.log(`Agent: ${result.finalOutput}\n`);
35
+ console.log(`Assistant: ${result.finalOutput}\n`);
36
}
37
rl.close();