A private memory for your project,
searchable from your editor.
Doable Brain is a Postgres and pgvector layer that sits next to one of your projects. It indexes your repo, your deploy logs, the names of your env vars, and any docs you upload. Claude Code, Cursor, and Codex then read from it over MCP, so they already know your codebase the moment you open a new chat.
14-day money back. Cancel any time. Your data, your keys.
Your editor forgets. Your Brain doesn't.
A coding agent opens every new chat from zero. The Brain is the stuff you wish it already knew: the layout of your repo, the decisions you wrote down last week, what's actually deployed. It stays in your account, and the next chat can read from it the moment it opens.
Indexed from what you already have
GitHub repos, deploy logs, the names of your env vars (never the values), project metadata, plus markdown and PDF you upload. Re-syncs only touch the files that actually changed.
Works in your editor
Run doable setup once. The brain_search and brain_remember tools show up automatically in Claude Code, Cursor, and Codex. No extra config.
Hybrid search
pgvector for semantic matches, pg_trgm for exact identifiers like function names and ticket numbers. Every query runs both and blends the rankings.
Bring your own keys
On the BYO tier, plug in your own OpenAI or Anthropic key. You pay your provider directly; we just bill for the storage and sync.
Ask in plain English
Pro adds an Ask button on the Brain page. It runs retrieval, calls your chosen model (gpt-4o-mini or claude-haiku-4-5), and returns an answer with citations back to the source.
Private by default
Every read and write is logged. Memories live in your project's row in our Postgres. They're never shared, never used to train anything, and you can export or delete the whole thing at any time.
Three steps, about two minutes.
- 1
Attach a Brain to a project
Pick Starter (we run the embeddings), Pro, or BYO (your own keys). We set up a per-Brain schema in our Postgres for you.
- 2
Add sources
Connect a GitHub repo, opt in to deploy logs, env names, and project metadata, or upload markdown, plain text, and PDF. We chunk, embed, and keep them in sync.
- 3
Search from your editor
Ask any question in Claude Code, Cursor, or Codex. The MCP tool returns the top chunks with citations back to the source. Save decisions back with brain_remember.
Brain pricing
Separate from your Cloud plan. Works on any project, free or paid, on Doable Cloud or your own server.
Brain Starter
One project, we run the embeddings.
- ✓Up to 50,000 indexed chunks
- ✓Weekly sync of your sources
- ✓We run the embedding calls
- ✓MCP brain_search and brain_remember
- ✓Dashboard search UI
Brain Pro
Bigger budget, hourly sync, Ask the Brain.
- ✓Up to 500,000 indexed chunks
- ✓Hourly sync
- ✓Ask the Brain with your OpenAI or Anthropic key
- ✓Provider-aware analyzer warnings
- ✓90-day audit log
Brain BYO
You bring the AI keys, we run the storage.
- ✓Up to 1,000,000 indexed chunks
- ✓Hourly sync
- ✓Your OpenAI or Anthropic key for embeddings and answers
- ✓You pay your AI provider directly
- ✓180-day audit log
Brain Enterprise
Cross-project, SSO, custom retention.
- ✓Account-scoped Brain across every project
- ✓Dedicated embedding endpoint
- ✓SSO and role-based access
- ✓Custom retention and data residency
- ✓Priority support
Doable Cloud hosting is billed separately. See doable.do/#pricing.
What your editor sees
Run doable setup once and two tools show up. We wrote the descriptions so the model knows when to use them on its own. No slash commands required.
// doable.brain_search
Search the project's Brain (the indexed memory layer) and return the top-matching chunks. Use this when the user asks "search my Brain", "look up X in my notes", "what did we say about Y", or any question that depends on the project's code, docs, deploys, env keys, or saved memories.
// doable.brain_remember
Store a durable note in the project's Brain. Use this when the user says "remember that …", "save this note", "add to memory", or when you (the agent) want to record a decision / insight / fact for future Brain searches to find.
Common questions
Do I have to host my app on Doable to use Brain?
No. A Brain attaches to a Doable project, but that project can be on your own server (BYO) or just a placeholder you set up to hold the Brain. You don’t have to move your hosting.
Where does the data live?
In our Postgres. Each Brain has its own row-scoped storage. We don’t hand Brain data off to third-party vector services, and your content is never used to train any model.
What does BYO mean in practice?
On the BYO tier you paste your own OpenAI or Anthropic key into the Brain settings. We encrypt it at rest and use it for embeddings and the Ask flow. Your AI provider bills you directly; we only charge for storage and sync.
How is this different from Mem, Letta, or Cursor’s memory?
Brain is per-project, it’s built on Postgres and pgvector, and it talks to your editor through MCP. The same Brain works in Claude Code, Cursor, Codex, and anything else that speaks MCP. The data lives next to your project, not in another vendor’s account.
Can my team share a Brain?
Yes. Anyone with access to the Doable project can read and write its Brain. Account-wide Brains that span multiple projects ship on the Enterprise tier.
Can I export it?
Yes. From the dashboard you can download every chunk, memory, and audit row as JSON. Your data is yours.
Give your project a Brain.
14-day money back. Cancel any time. Your keys, your data, your memories.
Start a Brain