Coding agent memory your whole toolchain shares

Cursor wrote the decision yesterday. Claude Code should not ask again today. ReadmeDB is coding-agent memory as durable markdown — plus optional distilled facts with hybrid_recall when you need both layers.

Files + memories (hybrid)

Durable truths live in markdown (write_file, patch_file, version diffs). Quick scratch facts use add_memory and graduate via promote_memory_to_file. Neither Mem0-style vectors nor a second brain app required.

Tools coding agents actually use

• search_files / fetch_docs_for_task — find context fast
• append_to_file — log decisions without rewriting whole files
• get_context_profile — orient at session start
• hybrid_recall — search files and memories together

Connect in one config block

{
  "mcpServers": {
    "readmedb": {
      "url": "https://app.readmedb.com/api/mcp",
      "headers": {
        "Authorization": "Bearer rdb_your_key"
      }
    }
  }
}