One source of product truth. Read by every agent. Governed by your team.
Decisions, tensions, standards, and work packages — connected, typed, and current.
Read the docs · Join the community — everyone using this is still learning how to use it well, including me.
You already have context files, docs, chat history, and agent memory. The problem isn't that your agents have no context — it's that nobody knows which parts are still true, how they connect, or what matters most.
Product Brain fixes that. It's a governed knowledge graph for your product — decisions linked to the tensions that caused them, standards linked to the work they constrain, insights linked to the work packages they inform. Your agents read it before they act. Your thinking finally compounds instead of scattering.
The cockpit — every mode answers a different thinking question
Bridge mode — what needs your attention right now
The first time it surfaces something you forgot you logged.
Not “your AI didn’t make a mistake.” Better than that.
You’re shaping a new work package. Your agent reads the Chain, then says:
“This contradicts STD-47, which you wrote three weeks ago — feature flags are required for user-visible changes. Want me to factor that in, or are you revisiting the standard?”
You’d forgotten the standard existed. The agent didn’t.
That’s the moment Product Brain earns its place. Not because the agent got smarter, but because your thinking from three weeks ago is still in the room.
Four ways it changes your day.
Stay Oriented
What needs my attention right now? Open work packages, active tensions, decisions that need judgment — in one view.
Bridge ModeExecute Clearly
What’s in flight, blocked, or next? Your agent reads the state of every work package before writing code.
Execution ModeLearn from Every Session
What shipped, what broke, what patterns keep coming back? Decisions and lessons are captured back to the Chain.
Review ModeAsk with Full Context
Chat with an AI that already knows your product — decisions, standards, architecture, what’s in progress.
Brain ChatPlus: Strategy & Roadmap views, governance signals, team workspace — see the full method
Your AI sees this before it writes a single line of code
One kernel. Every agent reads it.
client + next
agent CHAIN.md
How it works
Three moves. One shared graph.
- pb orient
Orient
Your agent loads workspace context before it writes a line.
- pb capture
Capture
Decisions become versioned entries with human IDs — not chat threads.
- pb search
Search
Query by meaning, not file paths.
Prefer the exact commands? Show CLI examples
Run these from your project (or the same ideas via MCP in Cursor / Claude Code):
❯ pb orient -b
# Product Brain
Workspace ws-… — Product Brain is healthy.
## Active work packages
WP-211 [work package · active] …
❯ pb capture "DEC: Use Tailwind for all prototypes"
✓ Captured DEC-342: …
❯ pb search "auth friction"
TEN-445 [tension · open] …Three ways to start.
Install the CLI
For builders working in a terminal.
npm install -g @productbrain/cli Read CLI docs →Connect via MCP
For builders in Cursor, Claude Code, or any MCP-enabled tool.
Open MCP setup Read MCP docs →Create a hosted workspace
For teams that want the Cortex UI and zero setup.
Create workspace →Free during beta
Whichever path you pick, you’re not figuring it out alone. Everyone using Product Brain is still learning how to use it well, including me.
Join the community →