The brain behind
your agent.

Where you actually work

Start Codex CLI in your brain repository and work through the agent. Serenity supplies the memory tools underneath: bringing in sources, retrieving context, and preserving records beyond a conversation. You can also integrate it into another system, with that system providing the interface.

The brain repository holds your personal context. Keep it separate from the Serenity source-code repository.

Give the agent context worth keeping

Notes, repository documentation, and email contain decisions and details a model cannot know on its own. Choose which sources the brain ingests and when they sync. Your agent can use that material through Serenity’s interfaces.

Keep the evidence with the answer

Full-text search works without a model. With extraction and composition configured, Serenity can provide claims and cited answers to the agent or calling system. Citations let you inspect the original support.

Own the memory behind the conversation

Your brain’s durable records are inspectable files in Git. You choose the backup location and model providers. Model-assisted features send relevant context to the provider you configure.

Carry your constraints into other systems

Serenity can check structured actions against recorded constraints, called precepts. An integrating system can use the result in its workflow. Free-text checking remains unverified; the documented structured-action contract defines what works today.

Two ways to use the same foundation

• Through an agent: run Codex CLI in the brain repository and give it access to the documented Serenity commands or MCP tools.
• Inside another system: integrate through the CLI, MCP, HTTP interfaces, or the read-only Go package. Your application owns the user experience.

Read the agent and embedding guide →

Inspired by gbrain

Serenity was inspired by and builds on Garry Tan’s gbrain. David Ndungu began the project after watching Garry Tan: Own Your Intelligence.

The idea that carried through: keep your context in your own custody and make it available to the agent doing your work. Serenity develops that foundation as a memory and context layer you can run and embed.

A possible future Brain app

We may create a Serenity Brain app in the future. It is a possibility, not an available product or a promised release. Today, Serenity runs behind your agent or inside another system. This website explains setup; its chat only answers questions about public documentation.