gbrain

gbrain

Gives the AI agent you already use a memory you control, storing facts with their sources and allowing corrections and withdrawal. It starts keyless with keyword retrieval and adds semantic search later.

Made by garrytan on GitHub. These notes are Mindket's, written from its README; the project's own files and license are in its repository.

About

What it is: GBrain gives the agent you already use a memory that you control. It stores explicit facts together with their sources, lets you correct or withdraw them, and serves the same memory to all your agents. It begins with memory that needs no key and finds notes by keyword. Search by meaning, synthesis and background enrichment come later, when you want them.

Who it suits: People who run an AI agent such as OpenClaw or Hermes and want it to remember things reliably. Teams that use several agents and want one shared memory can use it too.

Industry | Who uses it | How they use it | Example
Software development | Developers with coding agents | Keep memory skills beside their knowledge so each agent works from the same facts | A fact stored with its source, then corrected when it changes
AI agent operations | People running more than one agent | Make one memory available across their agents | Two agents reading the same stored notes
Teams | Groups that share knowledge | Let agents join an authorized catalog where only approved editors publish | A shared set of memory skills that only approved editors change

Skip it if you do not use an AI agent.

Where to use it: Wherever your agent already runs, on your own machine or in a hosted version.

• Your existing agent: the recommended route, which keeps its identity and saves memory inside its environment
• Your own machine: the project is free to self-host
• ChatGPT, Claude and Cursor: the optional paid hosted version connects to them
• A shared catalog: agents join an authorized one, and approved editors publish to it

Use case | Who | What they get back
Store a fact with its source | Anyone using an agent | A fact that points back to where it came from
Correct or withdraw a fact | Anyone using an agent | Memory that changes when the facts do
Share memory across agents | Teams | The same memory in each agent

For newcomers and experts: If you are new, think of GBrain as a notebook your AI agent can read and write. It keeps each fact next to where it came from, and you can fix or remove a fact later. A good first thing to try is the keyless start, which needs no API key and finds notes by keyword.

If you know the field, retrieval begins with keywords and grows into semantic search, synthesis and background enrichment as you switch them on. A new local brain places its memory skills in the folder that holds its content. Publishing to a shared catalog is limited to approved editors. Managed coding-agent installs add a native router that GBrain owns, and restarting the agent and checking real use are separate steps. A staged migration keeps your sharing choices and personal edits.

An agent saves a fact with its source
You correct or withdraw it
Other agents read the same memory
You add semantic search when needed

Good to know: Nothing here runs by itself. The kit holds notes and the location of the code, and you follow the project's own setup.

• License: MIT, free to self-host
• Cost to run: Self-hosting is free. A hosted version is optional and paid. The source names no API keys.
• Data it reaches: The facts and sources your agents write to the memory. The hosted version connects to ChatGPT, Claude and Cursor.
• Not for: A tool that runs on its own. It works inside an agent you already use.

On GitHub, 10 November 2026

• Stars: 32,187 (216 that day)
• Forks: 4,769
• Watching: 135
• Open issues and pull requests: 799
• Language: TypeScript
• License: MIT
• Trending: #6 repository of the day on GitHub trending (10 November 2026)

Links

• Code: https://github.com/garrytan/gbrain
• Releases: https://github.com/garrytan/gbrain/releases
• Issues: https://github.com/garrytan/gbrain/issues
• Star history: https://www.star-history.com/#garrytan/gbrain&Date

Get the code

git clone https://github.com/garrytan/gbrain.git

Then follow the setup in its README. Nothing in this kit runs by itself.