pica-langchain

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

PICA MCP Integration with LangChain

PICA MCP 与 LangChain 集成

PICA provides a unified API platform that connects AI agents to third-party services (CRMs, email, calendars, databases, etc.) through MCP tool calling.

PICA提供统一的API平台，通过MCP工具调用能力将AI Agent连接到第三方服务（CRM、邮件、日历、数据库等）。

PICA MCP Server

PICA MCP 服务器

PICA exposes its capabilities through an MCP server distributed as

@picahq/mcp

. It uses stdio transport — it runs as a local subprocess via

npx

PICA通过以

@picahq/mcp

包分发的MCP服务器提供其能力，采用stdio传输协议——它通过

npx

作为本地子进程运行。

MCP Configuration

MCP 配置

json

{
  "mcpServers": {
    "pica": {
      "command": "npx",
      "args": ["@picahq/mcp"],
      "env": {
        "PICA_SECRET": "your-pica-secret-key"
      }
    }
  }
}

Package:
```
@picahq/mcp
```
(run via
```
npx
```
, no install needed)
Auth:
```
PICA_SECRET
```
environment variable (obtain from the PICA dashboard https://app.picaos.com/settings/api-keys)
Transport: stdio (standard input/output)

json

{
  "mcpServers": {
    "pica": {
      "command": "npx",
      "args": ["@picahq/mcp"],
      "env": {
        "PICA_SECRET": "your-pica-secret-key"
      }
    }
  }
}

包：
```
@picahq/mcp
```
（通过
```
npx
```
运行，无需安装）
鉴权：
```
PICA_SECRET
```
环境变量（可从PICA控制台获取：https://app.picaos.com/settings/api-keys）
传输协议：stdio（标准输入/输出）

Environment Variable

环境变量

Always store the PICA secret in an environment variable, never hardcode it:

PICA_SECRET=sk_test_...

Add it to

.env

and load with

python-dotenv

请始终将PICA密钥存储在环境变量中，切勿硬编码：

PICA_SECRET=sk_test_...

将其添加到

.env

文件中，并使用

python-dotenv

加载。

Using PICA with LangChain

在LangChain中使用PICA

LangChain provides MCP client support via the

langchain-mcp-adapters

package. Always refer to the latest docs before implementing. See langchain-mcp-reference.md.

LangChain通过

langchain-mcp-adapters

包提供MCP客户端支持。实现前请务必参考最新文档，参见langchain-mcp-reference.md。

Required packages

所需依赖包

bash

pip install langchain-mcp-adapters langgraph langchain-anthropic mcp python-dotenv

bash

pip install langchain-mcp-adapters langgraph langchain-anthropic mcp python-dotenv

Before implementing: look up the latest docs

实现前：查阅最新文档

The

langchain-mcp-adapters

API has changed across versions (e.g.,

MultiServerMCPClient

is no longer a context manager as of v0.1.0). Always check the latest docs before writing code. See langchain-mcp-reference.md.

langchain-mcp-adapters

的API在不同版本之间存在变更（例如，从v0.1.0版本起

MultiServerMCPClient

不再作为上下文管理器使用）。编写代码前务必查看最新文档，参见langchain-mcp-reference.md。

Integration pattern

集成流程

Create an MCP client using
```
MultiServerMCPClient
```
with stdio transport pointed at
```
npx @picahq/mcp
```
Get tools from the client via
```
await client.get_tools()
```

Create a ReAct agent using

create_react_agent(model, tools)

from

langgraph.prebuilt

Stream or invoke the agent with your messages
Pass environment variables (
```
PICA_SECRET
```
,
```
PATH
```
,
```
HOME
```
) to the MCP client's
```
env
```
config

创建MCP客户端：使用
```
MultiServerMCPClient
```
，将stdio传输指向
```
npx @picahq/mcp
```
获取工具：通过
```
await client.get_tools()
```
从客户端获取工具

创建ReAct Agent：使用

langgraph.prebuilt

提供的

create_react_agent(model, tools)

创建

流式调用或普通调用：使用你的消息调用Agent
将环境变量（
```
PICA_SECRET
```
、
```
PATH
```
、
```
HOME
```
）传入MCP客户端的
```
env
```
配置

Minimal example

最小示例

python

from langchain_anthropic import ChatAnthropic
from langchain_mcp_adapters.client import MultiServerMCPClient
from langgraph.prebuilt import create_react_agent

model = ChatAnthropic(model="claude-haiku-4-5-20251001", streaming=True)

client = MultiServerMCPClient({
    "pica": {
        "command": "npx",
        "args": ["@picahq/mcp"],
        "transport": "stdio",
        "env": {
            "PICA_SECRET": os.environ.get("PICA_SECRET", ""),
            "PATH": os.environ.get("PATH", ""),
            "HOME": os.environ.get("HOME", ""),
        },
    },
})
tools = await client.get_tools()
agent = create_react_agent(model, tools)

python

from langchain_anthropic import ChatAnthropic
from langchain_mcp_adapters.client import MultiServerMCPClient
from langgraph.prebuilt import create_react_agent

model = ChatAnthropic(model="claude-haiku-4-5-20251001", streaming=True)

client = MultiServerMCPClient({
    "pica": {
        "command": "npx",
        "args": ["@picahq/mcp"],
        "transport": "stdio",
        "env": {
            "PICA_SECRET": os.environ.get("PICA_SECRET", ""),
            "PATH": os.environ.get("PATH", ""),
            "HOME": os.environ.get("HOME", ""),
        },
    },
})
tools = await client.get_tools()
agent = create_react_agent(model, tools)

Invoke

调用

result = await agent.ainvoke({"messages": [{"role": "user", "content": "..."}]})

Or stream events

或者流式事件

async for event in agent.astream_events({"messages": messages}, version="v2"): kind = event["event"] if kind == "on_chat_model_stream": content = event["data"]["chunk"].content # content may be a list of content blocks (Anthropic models) or a string

undefined

async for event in agent.astream_events({"messages": messages}, version="v2"): kind = event["event"] if kind == "on_chat_model_stream": content = event["data"]["chunk"].content # content 可能是内容块列表（Anthropic模型）或字符串

undefined

Important: Anthropic content blocks

重要提示：Anthropic内容块

When streaming with

astream_events(version="v2")

, Anthropic models return

chunk.content

as a list of content blocks, not plain strings. Always handle both:

python

if isinstance(content, list):
    text = "".join(
        block.get("text", "") if isinstance(block, dict) else str(block)
        for block in content
    )
elif isinstance(content, str):
    text = content

使用

astream_events(version="v2")

进行流式调用时，Anthropic模型返回的

chunk.content

是内容块列表，而非纯字符串，请始终同时处理两种情况：

python

if isinstance(content, list):
    text = "".join(
        block.get("text", "") if isinstance(block, dict) else str(block)
        for block in content
    )
elif isinstance(content, str):
    text = content

Checklist

检查清单

When setting up PICA MCP with LangChain:

langchain-mcp-adapters

langgraph

langchain-anthropic

mcp

are installed

```
PICA_SECRET
```
is set in
```
.env
```
```
.env
```
is loaded via
```
python-dotenv
```
(
```
load_dotenv()
```
at top of file)
MCP client uses stdio transport with
```
npx @picahq/mcp
```
```
PATH
```
and
```
HOME
```
are passed in the MCP client
```
env
```
config
```
MultiServerMCPClient
```
is NOT used as a context manager (API changed in v0.1.0)
Streaming handles both list and string content from Anthropic models
Tool events (
```
on_tool_start
```
,
```
on_tool_end
```
) are handled for UI rendering

使用LangChain配置PICA MCP时的检查项：

已安装

langchain-mcp-adapters

、

langgraph

、

langchain-anthropic

、

mcp

```
.env
```
文件中已配置
```
PICA_SECRET
```
已通过
```
python-dotenv
```
加载
```
.env
```
（文件顶部调用
```
load_dotenv()
```
）
MCP客户端使用stdio传输协议指向
```
npx @picahq/mcp
```
MCP客户端的
```
env
```
配置中已传入
```
PATH
```
和
```
HOME
```
未将
```
MultiServerMCPClient
```
作为上下文管理器使用（v0.1.0版本API变更）
流式调用逻辑已处理Anthropic模型返回的列表和字符串两种内容格式
已处理工具事件（
```
on_tool_start
```
、
```
on_tool_end
```
）用于UI渲染

Additional resources

额外资源

For LangChain MCP adapter docs and API details, see langchain-mcp-reference.md

LangChain MCP适配器文档和API详情参见langchain-mcp-reference.md