Skip to content
Corpus. Docs

What is Corpus?

Corpus is a shared, Git-free home for the markdown collection your AI agents need to do good work.

You write documents in a web editor — coding standards, architecture notes, API conventions, runbooks, product decisions — group them into an ordered collection, and point your agent at that collection over MCP. Every agent on that collection reads the same source of truth, and you update it without touching a repo.

The problem it solves

Section titled “The problem it solves”

If you use Claude Code, Cursor, or any MCP-capable agent, you already have a collection problem:

  • The rules your agent should follow live in scattered markdown files, Notion pages, Slack threads, and people’s heads.
  • Each project re-pastes the same conventions into CLAUDE.md / .cursorrules, and they drift apart.
  • Non-engineers who own a lot of that knowledge can’t easily contribute to files behind a Git workflow.
  • When the canonical doc changes, nothing tells the agents.

Corpus makes one canonical copy, editable by anyone on your team in a browser, and serves it to agents on demand. No repo, no copy-paste, no drift.

The model

Section titled “The model”

Three concepts, in order:

  1. Document — one markdown file. It has a stable slug, a title, and an append-only version history. Editing it in the browser creates a new version; nothing is ever silently overwritten.
  2. Collection — an ordered list of documents, assembled into a single markdown corpus. This is the unit an agent consumes. A document can belong to many collections.
  3. MCP endpoint — one URL per project. An agent that authenticates to it can list your collections and read their assembled corpus (and individual documents). It is read-only: agents consume collections, they never write them.

Tenancy is Organization → Project. You sign up, name your organization, and a default project is created for you. Everything — documents, collections, team, the MCP endpoint — lives inside that project.

Who it’s for

Section titled “Who it’s for”

The same person, wearing two hats:

  • Writing collection — anyone on the team, technical or not, who knows how the work should be done. No Git, no PR, just a markdown editor.
  • Consuming collection — anyone running an agent who wants it to follow that knowledge. One config line and the agent has it.

Where to go next

Section titled “Where to go next”