The knowledge base your AI reads and writes.
Under the rules you set. A shared, governed memory for your team and the agents working beside them — with search that actually finds the right thing.
The whole loop, in 100 seconds.
How your team and your agents read and write one governed memory — under the rules you set.
Three things no “wiki + AI” gives you.
The combination is the moat: native agent authoring, real server-side retrieval, and human-curated rules — all multi-tenant from the ground up.
Agents read and write it
Native MCP. Claude, ChatGPT, Mistral — or any MCP client — searches, drafts, and revises your canonical docs through real tools, with safe transactional edits it can roll back. Not snippets. The source itself.
Search that finds things
Server-side hybrid search fuses keyword and meaning with multilingual embeddings, down to the passage. Ask in plain language; get the right paragraph back — not ten blue links.
Rules keep it honest
Human-curated, binding conventions are surfaced to every agent before it writes — per project or team. Governance is a first-class primitive, so the attic stays ordered no matter who’s filing.
Your AI is brilliant. It just doesn’t know your company.
brainattic gives it your knowledge — built as you work, governed by your rules, and reusable the moment it’s in. Capture what your team and your agents learn, then ask it back: to find an answer, make a call, or act on it.
Audit & document your infrastructure
Keep an always-current picture of your systems, networks, and services — and ask “what’s running where, and why?” to get a grounded answer instead of a wiki spelunk.
Make decisions that stick
Record the call and the reasoning behind it, so months later anyone — or any agent — can ask “why did we choose this?” without digging through chat history.
Keep a trail of work & communication
Briefs, hand-offs, who-did-what and what’s still open — searchable in one place instead of scattered across inboxes and threads.
Build a reusable proposal & account library
Templates, standard clauses, references, and everything you know about a customer — ready to pull into the next bid.
Run operations from living runbooks
How-tos and incident records your agents can actually read and follow, kept current as things change.
One question, across everything. Because it all lives in one governed knowledge base, your AI can answer a question that spans your infrastructure, a product, a customer, and a proposal at once — grounded in what your company actually knows.
Built for humans and agents alike.
Everyone works against the same governed source of truth — people in the editor, agents over MCP — and retrieves the same answer.
Add brainattic to Claude, ChatGPT, and Mistral.
No marketplace hunt, no glue code. Add it as a custom connector and your whole team’s agents pick up the toolset — each person scoped to exactly their workspace.
- Add the connector
Your workspace owner adds brainattic as a custom connector in Claude, ChatGPT, or Mistral — one URL. - Everyone authenticates
Each member connects and signs in individually. Nothing leaks across tenants. - Agents get to work
Search, author, revise, and recall — the full toolset, governed by your rules.
Ask your own agent.
brainattic is built to be read by AI — so don’t take our word for it. Paste this into Claude, ChatGPT, Mistral, or whatever agent you use, and let it tell you what brainattic is and what it could do with your knowledge base.
Works with any MCP client. Your agent reads our public llms.txt — no signup, nothing to install.
Fill it. Govern it. Find anything.
Fill the attic
Author in a fast markdown editor — or let an agent draft and you approve. Every change is versioned and restorable.
Set the rules
Define the conventions agents must follow, per project or team. They read the rules before they write a single line.
Find anything
People and agents retrieve the same canonical answer through hybrid search — instantly, in any language.
Open the attic.
We’re onboarding a small number of teams. Tell us a little about yours and we’ll set up a pilot.
No spam. We’ll only use this to talk to you about a pilot.