About myBrain
myBrain is a personal knowledge base that turns uploaded, linked, recorded, or written material into a searchable knowledge graph. Its MCP server lets AI agents search and explore that data, add sources for approval, and synthesize cited answers.
Key features
- 01Search and explore a knowledge graph
- 02Recall memories and source passages
- 03Add text, URL, and YouTube sources
- 04Approve sources before processing
- 05Synthesize answers with citations
- 06Manage access scopes and revoke clients
Use cases
- 01Search personal material while working with an AI agent
- 02Explore links between concepts and memories
- 03Submit new sources for review and approval
- 04Get synthesized answers with citations
MCP support
The hosted MCP server is available over Streamable HTTP with OAuth 2.1 or bearer API keys. It exposes read, write, and orchestration tools, with writes held for approval.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://mcp.mybrain.ai/mcp/claude mcp add --transport http mybrain-ai https://mcp.mybrain.ai/mcp/{"mcpServers":{"mybrain-ai":{"type":"http","url":"https://mcp.mybrain.ai/mcp/"}}}Open in Cursor{"servers":{"mybrain-ai":{"type":"http","url":"https://mcp.mybrain.ai/mcp/"}}}Open in VS Code{"mcpServers":{"mybrain-ai":{"type":"http","url":"https://mcp.mybrain.ai/mcp/"}}}{"mcpServers":{"mybrain-ai":{"serverUrl":"https://mcp.mybrain.ai/mcp/"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
https://mcp.mybrain.ai/mcp/Auth requiredTools
Details
From the project READMEmyBrain MCP
myBrain is your personal knowledge base: everything you upload, link, record or write becomes a knowledge graph you can search, explore and reason over. This repository is the public home of the MCP server entry and of the files that teach an AI agent how to use your brain well.
| Endpoint | Transport | Auth |
|---|---|---|
https://mcp.mybrain.ai/mcp/ | Streamable HTTP | OAuth 2.1 sign-in, or Authorization: Bearer <MCP API key> |
Keep the trailing slash in the URL.
Clients that support OAuth (claude.ai, ChatGPT, Claude Code) need only the URL: they open the myBrain sign-in, you choose what to grant on the consent page, and the client receives its own token. Every other client uses an MCP API key, created in myBrain: Settings → Connected Apps → MCP Integrations.
Both carry scopes: mcp:read (search and read tools), mcp:write (add sources,
approve, import memories), mcp:orchestrator (ask_mybrain). Signed-in apps and API
keys are listed in the same Settings page, and revoking one there cuts its access
immediately.
Connect
claude.ai (also used by the Claude desktop and mobile apps): Settings → Connectors →
Add custom connector, paste https://mcp.mybrain.ai/mcp/ and sign in to myBrain.
ChatGPT (plans with Developer mode): Settings → Apps & Connectors → Advanced
settings → turn on Developer mode, then Create a connector with the URL
https://mcp.mybrain.ai/mcp/ and authentication OAuth, and sign in to myBrain.
Claude Code, signing in:
claude mcp add --transport http mybrain https://mcp.mybrain.ai/mcp/
claude mcp login mybrain # or /mcp inside a session
or with an API key:
claude mcp add --transport http mybrain https://mcp.mybrain.ai/mcp/ --header "Authorization: Bearer YOUR_MCP_API_KEY"
Claude Desktop (config file), Cursor, VS Code (mcpServers in the client config):
{
"mcpServers": {
"mybrain": {
"url": "https://mcp.mybrain.ai/mcp/",
"headers": { "Authorization": "Bearer YOUR_MCP_API_KEY" }
}
}
}
Teach your agent to use it
Connecting the server is not enough. The agent also has to know when to search your brain instead of asking you, which tool is the cheap one, and that every write becomes a source pending your approval. Pick what fits your client:
| File | Use |
|---|---|
skill/mybrain/SKILL.md | Claude Code skill. Copy the skill/mybrain/ folder into ~/.claude/skills/ (personal) or <project>/.claude/skills/ (project). |
templates/CLAUDE.md | Block to paste into a project or global CLAUDE.md. |
templates/AGENTS.md | Same guidance for clients that read AGENTS.md (Codex, Cursor, others). |
All three assume the server is connected as mybrain (tool names mcp__mybrain__*
in Claude Code).
Tools
Read (mcp:read): recall_memories, list_memories, search_user_knowledge,
search_passages, read_source_passages, search_knowledge, search_sources,
list_sources, explore_node, trace_path, get_knowledge_overview,
list_knowledge_fields, get_knowledge_node, get_related_concepts,
list_chat_threads, get_chat_history, get_suggested_questions,
get_ingestion_status, get_usage_summary.
Write (mcp:write): add_source_text, add_source_url, add_source_youtube,
approve_source, import_memories. Writes land as a source pending your approval in
myBrain and are processed only after you approve.
Ask (mcp:orchestrator): ask_mybrain, a full answer synthesized over your brain with
citations. It is rate limited and spends your plan quota.
Registry
server.json is the manifest published to the
official MCP Registry as
io.github.mybrain-ai/mcp.
Publishing (maintainers only) needs an Owner of the mybrain-ai org. The
interactive mcp-publisher login github only grants the personal namespace
(registry#1468); log in
with a PAT instead (classic with read:org, or fine-grained with "Organization →
Members → Read-only"), on mcp-publisher 1.8.1 or newer:
./mcp-publisher login github --token "$GH_PAT"
./mcp-publisher publish
A published version is immutable, so every change to server.json ships with a new
version.
License
The files in this repository are released under the MIT License. See
TRADEMARKS.md for the marks. The myBrain service and the server
code behind mcp.mybrain.ai are not part of this repository.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.