GBrain plugin
GBrain stores sourced facts, searches them and gives Claude Code, Codex and Cursor one shared memory the owner can correct.
- 32,817 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Background jobs read mail, meetings and messages and store them in the base.
- Shared access does not cover every path; local files sit outside the boundary.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/gbrainDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Background jobs read mail, meetings and messages and store them in the base.
- Shared access does not cover every path; local files sit outside the boundary.
This is third-party code. Review the repository files before installing.
What it does
GBrain, an agent memory by Garry Tan, stores explicit facts with their sources and supports corrections and withdrawal. One brain serves several agents: keyword search works without keys, while semantic search, cited synthesis and background enrichment are added when needed. Storage is Postgres or PGLite with pgvector plus a typed link graph, so a question such as who works at a company has an answer. Claude Code and Codex get a plugin with curated skills, the `gbrain serve` MCP server and a `gbrain setup` command that writes a use-the-brain rule into agent instructions.
Who it is for. People running several agents who want a shared knowledge base of people, companies and projects on their own hardware.
Good fit when
- Claude Code and Codex must remember project decisions across sessions.
- An agent builds person and company pages from emails, meetings and notes.
- An answer is needed with citations and a note on what the base lacks.
- A team shares one brain with grants per source and operation.
Not a fit when
- A `CLAUDE.md` file or a short project note is enough.
- You do not want to maintain Postgres, background jobs and upgrades.
- You need a guarantee that shared access leaks nothing: the author's tests cover specific paths only.
Example request
What does my brain know about this project? Answer with citations and list separately what the base lacks.Limitations
Bun 1.4.4 or newer is required. GBrain is not on npm and the unrelated `gbrain` npm package must not be installed. The author describes his own setup of 155 thousand pages with dozens of autonomous jobs; a home project needs far less. Connecting mail, calendar, voice and Slack sends data to external services under your keys. The access boundary differs for remote clients and local files, so read the sharing boundaries before opening shared access. Read the README installer before running it.
How to disable. Run `gbrain setup claude-code --no-instructions` to drop the rule from `~/.claude/CLAUDE.md`, then remove the plugin through `/plugin` and the `gbrain` MCP entry. Delete the database and keys separately.
FAQ
How do I connect GBrain to Claude Code?
Run `/plugin marketplace add garrytan/gbrain`, then `/plugin install gbrain@gbrain`. For a local MCP server use `gbrain setup claude-code`.
Is GBrain installed through npm?
No. The npm `gbrain` package is unrelated. Use `bun install -g github:garrytan/gbrain#latest-stable` or clone the repository.
Are keys needed to start?
No. Memory and keyword search work keyless. Keys are needed for semantic search, synthesis and enrichment.
What is the hosted gbrain.io?
Optional paid hosting from the author. GBrain itself is MIT licensed and free to self-host.
What goes to external services?
Only what you connected yourself: embedding provider keys, mail, calendar, voice. Without them memory and keyword search run locally.
Discussion
No questions or reviews yet.
Related
PluginsBrainDB keeps agent memory in PostgreSQL: entities, relations, search and rules, with two skills wiring it to Claude Code.
Knowledge-graph memory for agents: Claude Code and Codex plugins, an MCP server and a Python library
BrainAPI turns text into a knowledge graph on your server and serves agents over MCP, REST and a console with fact-path answers
Memory for AI agents with three types: facts, events and procedures that learn from failures; cloud by key or a local folder
Memory for agents over MCP: five tools, one local SQLite file and layers modeled on human memory, no account
An MCP server and API of shared memory for several agents: facts, decisions and events live in one database with semantic search.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| GBrain (this card) | no | API key | 32,817 | |
| Mem0 | no | 68,887 | a cloud memory layer; GBrain keeps sourced facts locally and builds a typed graph | |
| Supermemory Skills | yes | 117 | skills for the Supermemory service; GBrain is a self-contained base with search and synthesis |
README in short
The README walks from choosing a setup to installation, MCP clients, data sources, page schemas and architecture, with notes on access boundaries.