AgentMail 是一个面向 AI 智能体的收件箱即 API 平台。该 langchain-agentmail 包提供了 LangChain 工具，用于发送消息、在线程中回复、管理草稿、下载附件和标记——所有功能都连接到真实的 AgentMail 收件箱。

概述

集成详情

可用工具

设置

安装包：

pip install -qU langchain-agentmail

凭据

您需要一个 AgentMail API 密钥。在 agentmail.to 注册以获取密钥。

if not os.environ.get("AGENTMAIL_API_KEY"):
    os.environ["AGENTMAIL_API_KEY"] = getpass.getpass("AgentMail API key:\n")

实例化

使用工具包一次性获取所有工具：

from langchain_agentmail import AgentMailToolkit

toolkit = AgentMailToolkit.from_api_key()
tools = toolkit.get_tools()

您也可以直接实例化各个工具：

from langchain_agentmail import AgentMailSendTool, AgentMailReplyTool

send = AgentMailSendTool()
reply = AgentMailReplyTool()

调用

直接使用参数调用

from langchain_agentmail import AgentMailSendTool

tool = AgentMailSendTool()
result = tool.invoke({
    "inbox_id": "ib_abc123",
    "to": "alice@example.com",
    "subject": "Hello from LangChain",
    "text": "This message was sent through langchain-agentmail.",
})
print(result)

使用 ToolCall 调用

model_generated_tool_call = {
    "args": {
        "inbox_id": "ib_abc123",
        "to": "alice@example.com",
        "subject": "Follow up",
        "text": "Just checking in.",
    },
    "id": "1",
    "name": "agentmail_send_message",
    "type": "tool_call",
}
tool_msg = tool.invoke(model_generated_tool_call)
print(tool_msg.content)

在智能体中使用

from langchain_agentmail import AgentMailToolkit
from langchain.chat_models import init_chat_model
from langgraph.prebuilt import create_react_agent

model = init_chat_model(model="claude-sonnet-4-6", model_provider="anthropic")

toolkit = AgentMailToolkit.from_api_key()
agent = create_react_agent(model, toolkit.get_tools())

response = agent.invoke({
    "messages": [(
        "user",
        "Check my inbox for anything new. Summarize the most recent thread in 2 sentences.",
    )]
})

草稿

草稿工具让智能体能够迭代编写、修改和发布——当模型想要暂存消息并在发送前确认，或通过 send_at:

from langchain_agentmail import (
    AgentMailCreateDraftTool,
    AgentMailUpdateDraftTool,
    AgentMailSendDraftTool,
)

created = AgentMailCreateDraftTool().invoke({
    "inbox_id": "ib_abc123",
    "to": "alice@example.com",
    "subject": "Draft v1",
    "text": "First pass — will be revised before sending.",
})

# `created` is a JSON string; parse to get the draft_id

draft_id = json.loads(created)["draft_id"]

AgentMailUpdateDraftTool().invoke({
    "inbox_id": "ib_abc123",
    "draft_id": draft_id,
    "subject": "Draft v2 — ready",
    "text": "Revised body. Sending now.",
})

AgentMailSendDraftTool().invoke({
    "inbox_id": "ib_abc123",
    "draft_id": draft_id,
})

附件

出站附件可以作为 base64 文件内容或公共 URL 传递。入站附件通过预签名下载 URL 获取：

from langchain_agentmail import AgentMailSendTool, SendAttachmentSpec

AgentMailSendTool().invoke({
    "inbox_id": "ib_abc123",
    "to": "alice@example.com",
    "subject": "Report attached",
    "text": "See the attached file.",
    "attachments": [
        SendAttachmentSpec(
            filename="report.txt",
            content_type="text/plain",
            content=base64.b64encode(b"hello").decode(),
        )
    ],
})

API 参考

有关 AgentMail API 的详细文档，请访问 docs.agentmail.to。Python 包源代码位于 github.com/agentmail-to/langchain-agentmail.