Claude Code, Cursor, Windsurf, and every other AI agent you use start each session cold. ReadmeDB gives all of them a shared, cloud-hosted markdown file store — so every agent reads the same context, remembers the same conventions, and picks up exactly where the last one left off.
You've explained your codebase architecture dozens of times. You've described your naming conventions, your open tasks, your team's preferences. Every new session, every new AI agent, every new chatbot — you start over.
Local files help if you're on one machine with one agent. But as soon as you're switching between Claude Code, Cursor, and Windsurf — or as soon as you want a cloud agent to have the same context as your local one — local files break down completely.
What you need is a single, cloud-hosted store of markdown files that every agent can read and write, from anywhere, with no setup beyond pasting a config block.
ReadmeDB is a hosted MCP (Model Context Protocol) server — a standard protocol that every major AI coding tool already speaks. Your agents connect to it with one JSON config block and immediately gain access to a persistent, versioned, searchable markdown file store.
Every agent — whether it's Claude Code on your laptop, Cursor in your office, or a custom bot in the cloud — reads from and writes to the same files. Memory is shared. Context persists. You stop re-explaining your codebase.
ReadmeDB works with any host that supports the Model Context Protocol. That currently includes:
No credit card required. Create your account →
Each key is an isolated file namespace. Create one key per project, or share one key across all your agents — it's your call.
Paste this into your .claude/config.json, .cursor/mcp.json, or equivalent for your editor:
{
"mcpServers": {
"readmedb": {
"url": "https://app.readmedb.com/api/mcp",
"headers": { "Authorization": "Bearer rdb_your_key" }
}
}
}Every agent now has 16 tools: read_file, write_file, search_files, append_to_file, get_file_history, and more. They read and write to the same store. Context is shared across sessions, across editors, across machines.
Any markdown file your agent creates to help itself in future sessions. Common patterns:
CLAUDE.md / AGENTS.mdProject-level instructions your agent reads at the start of every session — architecture, conventions, rules.
memory.mdRunning notes your agent appends to as it learns things about your codebase during a session.
tasks.mdOpen tasks, half-finished work, and next steps — so the next session picks up exactly where this one left off.
decisions.mdArchitecture decisions with context and rationale — so your agent doesn't re-litigate the same choices.
conventions.mdNaming rules, code style, patterns the team uses — eliminates the briefing at the start of every session.
The free plan includes 50 files, 5 MB storage, 3 API keys, and 60 requests per minute — enough to get started with multiple agents. The Pro plan is £5/month with unlimited files, 50 GB storage, and unlimited API keys.
Start free — no card required →