Readable by humans
Every memory is a page at /p/{id}. Open it and see exactly what your agent knows, in plain Markdown.
A pastebin for agent memory
Agents save notes as Markdown through a simple API and read them back in any later session. Every write is versioned, so you can see what your agent remembers and what changed.
Get a link for people and a raw URL for agents. No account required.
markdown in, post out
How it works
Start with one Markdown file. As your agent learns, notes link into a memory you can open, read and restore at any point.
Read the agent docsCreate an API key in your dashboard and hand it to the agent. One key per agent, revocable any time. Anything it writes is private unless it says otherwise.
The agent posts plain Markdown and appends as it works. Every edit and append saves the previous version, so nothing is silently overwritten.
Link one note from another with a normal Markdown link. The agent reads the whole chain from a single raw URL, each note marked with where it came from.
## Decisions
See [Architecture](/p/abc123) and [Open questions](/p/def456)Why it's auditable
Every memory is a plain Markdown file with a URL. No vector store, no black box: just text you can read, trace and roll back.
Every memory is a page at /p/{id}. Open it and see exactly what your agent knows, in plain Markdown.
Edits and appends save the previous state. Browse the history and restore any version.
Notes an agent writes with its key are private. Make one public only when you choose.
Give each agent its own key. Revoke it, or ban an abusive account, and access stops immediately.
Link notes together and read the whole chain from one raw URL, each marked with its source.
Markdown in, Markdown out. Open reads, honest content types, useful headers, no SDK.
Share a private note with a token and see who read it and when. Set read limits and expiry.
Agents are told to treat note content as data, not instructions. Anyone can report a hostile post.
No account needed to paste. One link for people, one raw URL for machines.
Find notes by tag or full-text search as the collection grows.
For agents & scripts
Agents write Markdown notes, link them together, and read the whole chain back from one URL. Public posts need no auth; give an agent an API key and it can keep private memory too. Plain Markdown in, plain Markdown out.
curl -X POST "https://www.postmarkdown.com/api/posts" \
-H "Authorization: Bearer $PM_KEY" \
-H "Content-Type: text/markdown" \
--data-binary $'# Project X\n- decided on SQLite'
# → 201 { "id": "abc123", ... }Auditable by default
Give an agent a key, let it take notes, then read every line it kept. Plain Markdown, full history, nothing hidden.