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: Update LangChain provider page with create_agent pattern
Use create_agent/createAgent + invoke() pattern. Standardize to Steps pattern, persistent language tabs, API key callout, and consistent email task. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Sushmithamallesh committed
Feb 17, 2026 at 15:20 UTC
27ffc703c733853fe1ba26ff5f29a84908d85992
1 file changed
+44
-50
docs/content/docs/providers/langchain.mdx
+44
-50
@@ -1,14 +1,16 @@
1
---
2
title: LangChain
3
description: Build an agent with LangChain and enable it to use 800+ tools with Composio.
4
-keywords: [langgraph]
4
+keywords: [langchain]
5
---
6
7
-The LangChain Provider transforms Composio tools into a format compatible with LangChain's function calling capabilities.
7
+The LangChain provider transforms Composio tools into LangChain's [StructuredTool](https://python.langchain.com/docs/how_to/custom_tools/) format with built-in execution.
8
9
-## Setup
9
+<Steps>
10
+<Step>
11
+**Install**
12
11
-<Tabs items={["Python", "TypeScript"]}>
13
+<Tabs groupId="language" items={["Python", "TypeScript"]} persist>
14
<Tab value="Python">
15
```bash
16
pip install composio composio_langchain langchain langchain_openai
@@ -16,43 +18,54 @@ pip install composio composio_langchain langchain langchain_openai
18
</Tab>
19
<Tab value="TypeScript">
20
```bash
19
-npm install @composio/core @composio/langchain @langchain/openai @langchain/langgraph @langchain/core
21
+npm install @composio/core @composio/langchain langchain @langchain/openai
22
```
23
</Tab>
24
</Tabs>
25
+</Step>
26
24
-## Usage
27
+<Step>
28
+**Configure API Keys**
29
26
-<Tabs items={["Python", "TypeScript"]}>
30
+<Callout type="info">
31
+Set `COMPOSIO_API_KEY` with your API key from [Settings](https://platform.composio.dev/?next_page=/settings) and `OPENAI_API_KEY` with your [OpenAI API key](https://platform.openai.com/api-keys).
32
+</Callout>
33
+
34
+```txt title=".env"
35
+COMPOSIO_API_KEY=xxxxxxxxx
36
+OPENAI_API_KEY=xxxxxxxxx
37
+```
38
+</Step>
39
+
40
+<Step>
41
+**Create session and run**
42
+
43
+<Tabs groupId="language" items={["Python", "TypeScript"]} persist>
44
<Tab value="Python">
45
```python
46
from composio import Composio
47
from composio_langchain import LangchainProvider
31
-from langchain import hub # type: ignore
32
-from langchain.agents import AgentExecutor, create_openai_functions_agent
48
+from langchain.agents import create_agent
49
from langchain_openai import ChatOpenAI
50
35
-prompt = hub.pull("hwchase17/openai-functions-agent")
36
-openai_client = ChatOpenAI(model="gpt-5")
37
-
51
composio = Composio(provider=LangchainProvider())
52
+llm = ChatOpenAI(model="gpt-5.2")
53
54
# Create a session for your user
55
session = composio.create(user_id="user_123")
56
tools = session.tools()
57
44
-agent = create_openai_functions_agent(openai_client, tools, prompt)
45
-agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
58
+agent = create_agent(tools=tools, model=llm)
59
+result = agent.invoke({"messages": [("user", "Send an email to john@example.com with the subject 'Hello' and body 'Hello from Composio!'")]})
60
47
-agent_executor.invoke({"input": "Star the composiohq/composio repository on GitHub"})
61
+print(result["messages"][-1].content)
62
```
63
</Tab>
64
<Tab value="TypeScript">
65
```typescript
66
+// @noErrors
67
import { ChatOpenAI } from '@langchain/openai';
53
-import { HumanMessage, AIMessage } from '@langchain/core/messages';
54
-import { ToolNode } from '@langchain/langgraph/prebuilt';
55
-import { StateGraph, MessagesAnnotation } from '@langchain/langgraph';
68
+import { createAgent } from 'langchain';
69
import { Composio } from '@composio/core';
70
import { LangchainProvider } from '@composio/langchain';
71
@@ -64,39 +77,20 @@ const composio = new Composio({
77
const session = await composio.create("user_123");
78
const tools = await session.tools();
79
67
-const toolNode = new ToolNode(tools);
68
-
69
-const model = new ChatOpenAI({
70
- model: 'gpt-4o-mini',
71
- temperature: 0,
72
-}).bindTools(tools);
73
-
74
-function shouldContinue({ messages }: typeof MessagesAnnotation.State) {
75
- const lastMessage = messages[messages.length - 1] as AIMessage;
76
- if (lastMessage.tool_calls?.length) {
77
- return 'tools';
78
- }
79
- return '__end__';
80
-}
81
-
82
-async function callModel(state: typeof MessagesAnnotation.State) {
83
- const response = await model.invoke(state.messages);
84
- return { messages: [response] };
85
-}
86
-
87
-const workflow = new StateGraph(MessagesAnnotation)
88
- .addNode('agent', callModel)
89
- .addEdge('__start__', 'agent')
90
- .addNode('tools', toolNode)
91
- .addEdge('tools', 'agent')
92
- .addConditionalEdges('agent', shouldContinue);
93
-
94
-const app = workflow.compile();
95
-
96
-const finalState = await app.invoke({
97
- messages: [new HumanMessage('Star the composiohq/composio repository on GitHub')],
80
+const model = new ChatOpenAI({ model: 'gpt-5.2' });
81
+
82
+const agent = createAgent({
83
+ model,
84
+ tools,
85
+});
86
+
87
+const result = await agent.invoke({
88
+ messages: [{ role: "user", content: "Send an email to john@example.com with the subject 'Hello' and body 'Hello from Composio!'" }],
89
});
99
-console.log(finalState.messages[finalState.messages.length - 1].content);
90
+
91
+console.log(result.messages[result.messages.length - 1].content);
92
```
93
</Tab>
94
</Tabs>
95
+</Step>
96
+</Steps>