Markdump saves the important bits to a markdown library your assistants share: notes, decisions, runbooks, and context that Claude, Codex, ChatGPT, or anything with MCP can pick up later.
Close the tab and it’s gone. Switch from your phone to your laptop, or from Claude to ChatGPT, and you start over. Markdump gives your assistants one shared memory that outlives any single conversation: a library any agent can read, search, and build on.
Whatever your agents touch can live here as plain files you can open too.
Notes, ideas, and bookmarks in one place every assistant can reach and update.
Specs, decisions, and TODOs that Claude writes and Cursor or Codex read. No copy-paste between tools.
Ops runbooks and how-tos your agents keep current and query on demand.
Capture sources and summaries, then ask any assistant to synthesize across them.
Put Claude routines and Codex automations to work on a schedule: have them pull from the connectors those platforms already offer, like Slack, Gmail, Drive, and GitHub, then distill it all into your library. Wake up to a synthesized digest every assistant can build on.
No password to invent. We send a six-digit code and you’re in.
Add Markdump as a connector. That’s the entire setup.
New to MCP? It’s the standard way assistants connect to outside tools, like a USB port for your AI.
Then open your library and watch the file appear. Your agent just wrote to your folder, and the next one can read it.
Connect any MCP client. They all work the same library, so what one writes, the next one reads.
Plain .md files in folders. No proprietary format, no lock-in. Browse, search, or export the whole thing anytime.
Full-text grep, glob find, and semantic search across everything, for your agents and for you.
Agents make surgical changes by matching exact content context instead of rewriting whole files. Every write is versioned.
Markdump speaks OKF, the Open Knowledge Format: a light convention that turns a folder of notes into a map your agents can navigate. A few lines of frontmatter say what each file is and how it connects, so an assistant finds the right note and keeps it current instead of guessing. New agents learn the format the moment they connect.
Discoverable. Every file declares its type and a one-line summary, so search and agents surface the right thing.
Connected. Cross-links spell out how concepts relate, so an agent can follow the thread.
Self-describing. Reserved index.md maps and a log.md history give a table of contents and a paper trail.
Works with the assistants you already use
Markdump is a markdown library your AI assistants share. Claude, ChatGPT, and any MCP client can read, search, and write the same notes, and because it's just a folder of files, you can open and edit it yourself anytime.
Any client that supports MCP (the Model Context Protocol), including Claude, Claude Code, ChatGPT, Codex, and Cursor. They all connect to the same library, so what one assistant writes, the next one can read. Our ChatGPT app is pending review, so for now ChatGPT needs Markdump added manually as a connector through developer mode.
MCP (Model Context Protocol) is the standard way AI assistants connect to outside tools and data, like a USB port for your AI. Markdump exposes your file library to assistants over MCP.
Sign in with your email, then add your Markdump MCP endpoint (https://markdump.com/mcp) as a connector in your assistant. Then ask it to read or save a note and watch the file appear in your library.
Markdump is free for now: 20 MB free, 1000s of md files. We have not introduced paid plans yet, but paid plans are planned for the future. If you need more storage or have a specific use case, contact support.
Yes. Markdump stores plain .md files in folders with no proprietary format and no lock-in. You can browse, edit, or export everything anytime. Deletes are recoverable and every overwrite keeps a previous version.
Sign up free, connect your first assistant, and your next chat picks up right where the last one left off.