Use case · Knowledge base
An MCP knowledge base for your agents
Agents are only as good as the context they can reach. Writespace turns your documents into a queryable knowledge base your agents connect to over the Model Context Protocol — so the same notes, specs and decisions are available to every client, every session.
Free up to 10 documents — no card required.
Built for retrieval
Full-text search is ranked by Postgres ts_rank and returns highlighted
snippets, so you can drop Writespace straight into an agent as RAG retrieval with no
extra pipeline. Your model asks a question, gets the relevant passages, and works from
real source material instead of guessing.
Structured, queryable metadata
Every doc carries JSONB metadata your agents can shallow-merge and query by
containment — treat it as frontmatter your model can read and write. Ask for
everything with status: "accepted", or tag docs by topic and pull them
back on demand. The knowledge base stays organised because the agent helps keep it
organised.
Write-back, not just read-only
A knowledge base that only reads goes stale. Here, agents append findings, create new docs, and update sections as work happens — so the base reflects the latest state, not last month's. Writes broadcast over Realtime, and any open editor reloads instantly.
One base, every client
Connect Claude, ChatGPT, Gemini and Cursor to the same workspace at the same time. They share one source of truth instead of each keeping a private, divergent copy. Add teammates and the knowledge base is shared with people too, edited live with cursors.
Simple to wire up
It's a remote, spec-compliant MCP server. Issue a scoped access token, paste it into your client, and your knowledge base is online. Stable doc IDs mean links between docs survive renames — references your agent can rely on.
Start free with up to 10 documents. Scale to an unlimited knowledge base with Pro at $5 per user / month.