We love simplicity. Agents love markdown.Born in 2026. No MCP bolted onto a pre-AI tool.
Shared memory your team’s AI reads and writes — and you can audit.
A memory per client. Hand it over when the project ends.
Your whole team’s Claude Code knows what the team decided.
Company memory for AI, in your own AWS account, with Microsoft Entra sign-in and an audit trail.
MCP · API · Web
· hosted in Frankfurt (eu-central-1)
Who writes to a bundle
- MA@maraowneragent/archivistClaude Code
- JO@jonasowneragent/pagerCI job
- NO@norareads
People
Share a bundle with a person, read only or read and write. They work in the web app, on desktop or phone.
Agents
Claude, ChatGPT, Cursor or a CI job, connected over MCP. Each works for one person, with only the paths and rights that person gives it.
Every change
Stamped with who wrote it: @mara via archivist, or @jonas in person. Each write names the revision it read, so nobody overwrites blind.
Everything lives in bundles.
A bundle is a folder of memory for one subject, private or shared. Open one to see what it holds.
Simple to start. More useful the more it holds.
You shape each bundle yourself: its folders, its kinds of link, which agent may touch what.
Documents
Technical documentation, decisions, runbooks, meeting notes. Written by people, kept current by their agents.
adr/runbooks/diagrams/
Processes
A merge, an alarm or a schedule starts an agent. It reads the bundle, does its step and leaves the next one for a person.
incidentsonboarding/reviews
Results
What the work produced stays, with who made it and when: reports, findings, experiments.
reports/findingsexperiments/
Three bundles at work.
Examples built in Korpus; the screens are the real app. Humans and agents work in the same bundle at the same time.
Give your agents a memory your team can read.
Free to start, no card needed. For a demo or running it in your own AWS account, write to us.