What it does
Open Brain collects thoughts from a Telegram bot, from automated pipelines that ingest RSS feeds, Hugging Face daily papers and Emergent Mind, or from direct capture. Each thought is embedded, checked for semantic duplicates and linked to related thoughts, with connections above 0.80 similarity classified into typed relationships. Named entities are resolved into a shared graph, and a background Dream cycle deduplicates near-duplicates, tracks theme drift, synthesizes insights across clusters and archives stale content. The server exposes 17 MCP tools.
It is intended for MCP-compatible assistants such as Claude or ChatGPT that need to search a personal store of notes by meaning, filter it by type, topic, person or time, retrieve aggregate statistics, or request a weekly review of themes. The publisher reports 37.2% on a LongMemEval baseline.
The server runs either on Supabase, which is hosted, or on Docker Compose, which is self-hosted, with scheduling through GitHub Actions or a built-in cron container respectively. Setup uses bootstrap, deploy and validate scripts; validate runs eight checks against the live deployment. Once deployed, an AI client connects with an MCP endpoint URL and a generated access key.
Limitations and requirements
Requires an OpenRouter account and API key, used for generating embeddings and extracting metadata. The hosted option requires a Supabase account, the Supabase CLI, a project URL and a service role key. The Telegram bot is optional. Docker Compose deployment runs on your own machine or server.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 13 Nov 2026 · its page on the source
Open Brain
Graph-structured MCP memory server. 37.2% LongMemEval. Auto dedup, themes, decay, synthesis.
v1.1.0As shown on registry.modelcontextprotocol.io when we last read it