Email for LangChain
An email inbox for your LangChain agent
langchain-pidgeon gives a LangChain or LangGraph agent its own address, with nine tools to read, reply in the thread and wait for verification codes.
The toolkit reads PIDGEON_API_KEY and hands the model every tool. Each tool is one call of the Python SDK, so a refusal is the API’s own.
pip install langchain-pidgeonWhat your agent can do
- ToolsNine toolsThe hosted MCP server’s tools by name, so the model sees the same names however it reaches Pidgeon.
- CodesWait for a verification codeOne call waits for the next email and returns its code or link already pulled out.
- CleanBodies arrive cleanedHidden text removed and fake chat turns marked, before the model reads them.
- ApprovalAsk before it sendsEach tool says whether it is read-only, so LangChain’s human-in-the-loop middleware can stop at every write.
- KeysOnly what you grantedA read key cannot send, whatever the prompt says. The key is the whole of the access control.
- GraphLangGraph tooThe same toolkit works in a LangGraph agent.
Asked often
Is there a LangChain toolkit for email?
Yes: langchain-pidgeon, nine tools for a LangChain or LangGraph agent to read, search, reply and wait for verification codes in its own inbox.
How does the agent get a verification code?
The await tool holds one request open until mail arrives, 50 seconds by default and at most 280, and returns the codes and links already extracted.
Can a person approve what it sends?
Yes. Each tool carries hints saying whether it is read-only, so LangChain’s human-in-the-loop middleware can stop at every tool that is not.
What does it cost?
The free plan has unlimited inboxes, one custom domain and 100 emails a day, with no card. Receiving is free on every plan.