Provides large language models secure, tool-based access to email accounts over IMAP.
Sponsored
LLMail functions as an MCP server, empowering Large Language Models to interact with your email inbox through a secure, tool-based interface over IMAP. It supports multi-account IMAP, offers advanced Gmail features like X-GM-RAW search and labels, and includes a local full-text search index for efficient retrieval. A key focus is safety, with features like a CPU-based prompt injection guard and strict limitations preventing LLMs from sending emails or permanently deleting data. This ensures agents can manage messages and folders without compromising data integrity. Additionally, it features a built-in chat client for direct interaction with configured LLM providers like Anthropic or OpenAI.
Key Features
01Multi-account IMAP with connection pooling
02Native Gmail extensions for X-GM-RAW search and label support
03Local full-text search via Bleve with background sync
04CPU-based prompt injection guard using Meta's Prompt Guard 2
05Secure credential storage via system keyring or encrypted config
060 GitHub stars
Use Cases
01Enable AI agents to safely read, search, and manage email content and attachments.
02Integrate email capabilities into MCP-compatible LLM clients like Claude Desktop or Claude Code.
03Utilize a standalone chat client for LLM-powered email interaction and organization without external MCP hosts.
