MCP-ready project memory
Project memory your AI agents can trust.
Stop losing context between agent sessions. Store decisions, debugging trails, handoffs, and project facts in one fast workspace built for Codex, Claude, ChatGPT, and local agents.
{
"mcpServers": {
"knownbase": { "url": "http://127.0.0.1:3579/mcp" }
}
}
What agents store
The context that usually evaporates. Kept.
Code lives in git. The reasoning around it lives in Knownbase, so the next session doesn't start from zero.
Decision records
Why you chose this library, this schema, this trade-off, captured once so no agent relitigates it later.
Debugging trails
The investigation that finally explained a flaky test, so the fix and its cause survive past the session.
Session handoffs
What's done, what's in progress, and the one next step, ready for the next agent to pick up. How to write one →
Architecture notes
The mental model of the system: how the pieces fit, what the invariants are, where the landmines hide.
Project facts
Conventions, endpoints, credentials-to-rotate, gotchas: the small durable facts an agent needs on hand.
Team knowledge
Share one workspace so every teammate's agent reads and writes the same memory. No more per-person silos.
How it works
Three steps to memory your agents can trust.
Add the Knownbase MCP endpoint and an API key to Codex, Claude Code, ChatGPT, or any local agent.
Agents call upsert_note to save decisions, debugging trails, and handoffs into projects.
Any agent searches the workspace with search_notes and picks up exactly where the last one left off.
Affordable plans
Start free, upgrade when your agents need more memory.
Every paid plan starts with a 7-day free trial. No charge until it ends, cancel anytime.
FAQ
Built for how agent teams actually work
What is Knownbase?
Knownbase is MCP-ready project memory for AI agents. It keeps source code in git and working memory: decisions, debugging trails, architecture notes, and task handoffs, all in one fast, searchable workspace your agents read and write over the Model Context Protocol.
Which agents and tools work with it?
Anything that speaks MCP: Codex, Claude Code, ChatGPT, Cursor, and local agents. You add the Knownbase endpoint plus an API key, and the agent gets tools like search_notes and upsert_note.
Is my data isolated and secure?
Every note is scoped to your workspace and never shared across tenants. Sessions are hashed, API keys are stored hashed, and passwords use PBKDF2. You can export or delete your data at any time.
Can I start for free?
Yes. The Free plan includes 2 projects, 100 notes, and an MCP key. No card required. Upgrade when your agents need more memory.
How do I get support?
Email hello@knownbase.dev or use the contact form. We're happy to help.
From the blog
Learn how agent memory actually works.
What is the Model Context Protocol?
A plain-English guide to MCP, the standard way agents reach tools and data.
Why AI agents need persistent memory
Why agents forget between sessions, and what a real memory layer looks like.
Give Claude Code & Codex memory over MCP
A ten-minute setup walkthrough for connecting any MCP client to a memory server.