Agents propose, people approve, and every value keeps its source. The next agent and the next colleague start from what your team already knows. On your own server.
AI agents now draft documents, fill tables and answer questions all day. Without a shared, checkable memory, their work turns into noise.
What an agent found lives in one person’s chat. The next agent, and the next colleague, start from zero.
A value in a doc or a table has no source. When it is wrong, you can’t tell which document, agent or person it came from.
An agent with write access can overwrite a page or a record, and the only defence is a prompt asking it to be careful.
Most memory tools run in someone else’s cloud. Teams with security requirements can’t send them their documents.
An agent connects over MCP and reads the rules of the space. A write that breaks a required rule is refused. Where you ask for review, the agent leaves a proposal with its source. Once a person approves it, the value settles with its thread, and every agent that reads it later gets the provenance too.
Pages, databases and permissions, so your people can read, fix and approve what agents wrote. Not a black box behind an API.
Especially teams whose documents can’t go to someone else’s cloud. Sedimenta is in private preview with a few teams like these.
Teams of 10 to 200 that already work with Claude Code, Codex or Cursor: one memory of decisions, requirements and docs instead of markdown files scattered across repositories.
Agents read reports, websites and emails; analysts approve what goes into the market map, with a source on every number.
Policies, owners and review dates in one place, with a verifiable journal of who changed what and when.
This year our AI agents started doing real work: researching markets, drafting documents, writing code. They were fast — and forgetful. What one agent found was lost in a chat; the next one started from zero, or quietly overwrote a page with something it had made up.
We didn’t need a smarter model. We needed a place where agents and people share one memory: where every value shows its source, where rules are checked by the server instead of requested in a prompt, and where a person decides what becomes true for the team.
So we built Sedimenta — and we build Sedimenta inside it. Our whole development plan, more than a thousand items with every decision and review, lives there, and our agents work in it every day under the same rules. That is the memory we want to give your team.
— The Sedimenta team
We’re working with a small number of teams that already use AI agents in their daily work.