The company brain · for teams & their agents
Your team's memory,recalled.
Cerebro captures every decision, learning, and correction from the tools you already use — then hands it back, dated and sourced, the moment you or your AI agent asks. Stop re-deciding what you already decided.
MIT core · MCP-native · DPDP-compliant · exports to markdown
↑ our demo brain. ask it anything — then picture yours.
in Cerebro's own brain, since May — dogfooded daily
See it in 60 seconds
What Cerebro actually does
what your team forgot this week
Every real decision leaves a hole where it should have been kept.
slack · #architecture · 4,000 messages deep
Why we moved off the monolith
Debated for a week. The thread scrolled away. The reasoning left with the debate.
linear · API-482 · closed
The exact rate-limit numbers
Chosen deliberately. The engineer who chose them is on another team now.
google doc · shared with 2
The customer-success playbook
One of the two left. The other forgot the doc exists.
one brain, five kinds of memory
Not notes. Typed memory your agents can reason over.
Decisions
“Redis Cluster over single-node — we need cross-region.”
Learnings
“The outage root cause: pool exhausted at 200 conns.”
People
“Priya owns rate-limits; escalation path in onboarding.”
Todos
“Ship the Notion connector before Linear.”
Observations
“Enterprise leads convert 3× faster with SSO.”
memory over time
Ask what's true. And when.
When a decision reverses, the old entry doesn't vanish — it's stamped superseded and kept in the ledger. Your agent recalls today's answer, and can still show you March's. Nothing else in your stack does this.
“Cache in Memcached — it's already in our stack.”
“Redis single-node. Memcached can't do our eviction policy.”
“Redis Cluster — the single node fell over cross-region.”
nobody has to write it down
The record files itself.
Connectors watch the rooms where decisions actually happen. A resolved thread, a merged PR, a closed ticket — each one typed, dated, linked, and filed automatically. No new habit to build. Watch it arrive →
- slack#architecturedecisionjust now
- githubPR #1284 mergeddecisionjust now
- linearAPI-482 closeddecision1 min ago
- slack#incidentslearning2 min ago
- webhookdeploy.successobservation2 min ago
- githubissue #1301 openedtodo3 min ago
what it costs to keep the record
Free to feel it. Cheap to run it.
Pricing
Per-seat. Free for personal use, cheap for teams, custom for compliance. USD primary, ₹ PPP for India.
The memory you can also make it forget.
When a person asks to be forgotten, one action removes every entry that mentions them — the search index included — and logs the request. DPDP and GDPR aligned from day one.
On the record.
Is this just a wiki with extra steps?+
A wiki is written by hand and rots the day someone stops maintaining it. Cerebro is written by your connectors, typed by kind, dated, and superseded automatically — and it answers questions, from your AI agent or a search box, instead of waiting to be read.
How do our AI agents actually use it?+
It's MCP-native. Point Claude Code, Cursor, Windsurf, or any MCP client at your Cerebro endpoint and they get eight memory tools — read, search, write, supersede, and more. Your agents stop re-deriving decisions the team already made.
Where does our data live?+
Encrypted at rest on Supabase in India (ap-south-1). Per-subject erasure for DPDP and GDPR. No training on your data. Export to markdown any time — nothing holds you hostage. Or self-host the MIT core entirely.
Give your team a memory.
Three seats, one connector, and every MCP tool — free, forever. Your brain starts filling itself the moment you connect it.