Local search over Bitrix24 chat history
bitrix-chat-mcp
An MCP server with a local SQLite mirror of Bitrix24 chat history: full-text search, forgotten questions and unkept promises, zero writes back to Bitrix
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- It stores conversations, including personal dialogs, on disk in plain form
- The im-scoped webhook grants chat read access under the rights of the user who created it
Install
Manual install
make install && cp .env.example .env && make ui && make syncInstall, set up the webhook via .env, pick chats through the UI, and run the first sync.
This is third-party code. Review the repository files before installing.
What it does
The server pulls the history of selected Bitrix24 chats into a local SQLite database with an FTS5 full-text index, exposing tools to search and analyze it: find a message by topic, see the surrounding context, bookmark an agreement, find unanswered questions (by topic keywords from config.json) and your own unkept promises. A web UI for choosing which chats to mirror listens only on 127.0.0.1 and never displays or accepts the webhook itself. The server never writes back to Bitrix24; its only write is a local bookmark in its own database, and it is designed to pair with a regular Bitrix24 MCP server that shares the same dialog and message identifiers.
Who it is for. For teams whose important technical agreements get buried in Bitrix24 chat history and are hard to search through the live API.
Good fit when
- You need to quickly find where a specific decision or remark was discussed
- You need to track questions you have not answered or your own unkept promises
- Privacy matters: data should stay on the local machine, no cloud, no embeddings
Not a fit when
- You need to send messages to Bitrix24 chats: this server only reads; for writing you need bitrix24-local-mcp
- Security requirements forbid storing personal conversations on disk in plain form
Example request
Find where the ai-rules-check rules were discussed in chats and show the context around that messageLimitations
It never writes to Bitrix24, only local bookmarks. Files and attachments are not downloaded, and system messages and calls are not indexed. Detecting kept promises and answered questions relies on a rough heuristic, whether you wrote in the chat afterward, not an exact match to the original message. Mirroring stores conversations on disk in plain form, including personal dialogs, unless explicitly excluded.
How to disable. Remove the bitrix-chat entry from your MCP client config and delete the data/ folder with the local mirror database.
MCP
- Transport
- stdio
- Authentication
- API key
| Environment variables | |
|---|---|
| BITRIX24_WEBHOOK_URL required, secret | Bitrix24 inbound webhook with im and user rights, read from .env or a sibling bitrix24-local-mcp/.env |
Security check
- It stores conversations, including personal dialogs, on disk in plain form
- The im-scoped webhook grants chat read access under the rights of the user who created it
README in short
The README compares the server to bitrix24-local-mcp in a table (live state vs. accumulated memory), describes the shared credential and identifier vocabulary between them, install via make install/ui/sync, how the answer_pending and my_promises tools work with nuances worked out on real data (372 messages over two weeks), and the project layout.
FAQ
Can the server post a message to Bitrix24?
No, it never writes to Bitrix; its only write is a local chat_pin bookmark in its own database.
How does it relate to a regular Bitrix24 MCP server?
They share a dialog_id and message_id vocabulary: a message found here can be opened right in the live MCP to reply.
Related
A self-hosted knowledge base with block-level references and a built-in MCP server for connecting AI agents to your notes
A CLI for every Google Workspace API with JSON output and agent skills: Drive, Gmail, Calendar, Sheets and more
Local search over Markdown notes, docs and meeting transcripts: keywords, semantic search and reranking, with an MCP server
A task manager for AI-driven development: breaks a PRD into dependent tasks and guides the agent through them via MCP or CLI