Connect
Execution memory for your AI agents
Connect Claude, ChatGPT, Codex or any MCP client and your agent can resume where you left off, capture decisions as it works, and keep each pad's rituals current.
MCP endpoint
https://next.nodepad.tools/mcp
Add it as a custom connector. Sign-in happens through OAuth in your browser; read-only connections are supported.
Claude Code
claude mcp add --transport http nodepad https://next.nodepad.tools/mcp
Codex
In ~/.codex/config.toml, with the RMCP OAuth client and Nodepad's exact resource:
experimental_use_rmcp_client = true
[mcp_servers.nodepad]
url = "https://next.nodepad.tools/mcp"
oauth_resource = "https://next.nodepad.tools/mcp"
Restart Codex, run codex -c experimental_use_rmcp_client=true mcp login nodepad, check the browser URL contains resource=, then start a fresh session.
The np CLI
Create a token under Sign-in & API tokens, then point the CLI here:
export NODEPAD_API_URL=https://next.nodepad.tools
export NODEPAD_TOKEN=np_…
np resume
What your agent can do
nodepad_workspace— Nodepad Workspacenodepad_node— Nodepad Nodenodepad_capture— Nodepad Capturenodepad_project— Nodepad Projectnodepad_binding— Nodepad Bindingnodepad_history— Nodepad Historynodepad_runbook— Nodepad Runbook
Why it's safe
- OAuth with PKCE; tokens are bound to this server and work nowhere else.
- Removal and unbinding need explicit confirmation.
- The
mcp_readscope gives an agent read-only access.