todo-mcp-http

The same todo MCP server as todo-mcp, but served over Streamable HTTP on localhost instead of stdio — demo for the talk MCP — Plug Your Agent Into Everything (FIT-HCMUS).

Same capabilities: tools add_task · list_tasks · complete_task, resource todo://list, prompt plan_my_day — plus a REST API over the same state (see below).

Why this variant exists — the state lesson

	stdio (`todo-mcp`)	HTTP (this)
Who starts the server	each host spawns its own child process	you start one process, hosts connect to it
Processes	one per host	one, shared
State	separate list per host, dies with the session	one list shared by every client, lives until you stop the server

Two Claude Code windows connected to this server see (and edit) the same todo list. With the stdio variant, each window has its own.

Sessions still exist: each client gets its own Mcp-Session-Id (protocol bookkeeping), but the tasks array is module-level — one per process.

Run

npm install
npm start          # → http://localhost:3001/mcp  (PORT=xxxx to change)

Test with the MCP Inspector

npx @modelcontextprotocol/inspector

In the UI choose Streamable HTTP and enter http://localhost:3001/mcp.

Plug into Claude Code

claude mcp add --transport http todo-http http://localhost:3001/mcp

Open two terminal windows running claude, add a task in one, list in the other — shared state, live.

REST API — the same state, a second door

Three endpoints mirroring the three MCP tools, on the same server & state:

curl http://localhost:3001/api/tasks                          # list
curl -X POST http://localhost:3001/api/tasks \
  -H 'Content-Type: application/json' -d '{"text":"hello"}'   # add → 201
curl -X POST http://localhost:3001/api/tasks/1/complete       # complete

Add a task with curl, then ask Claude to list_tasks — it’s there. REST is the door for humans and scripts; MCP is the door for models. Same operations underneath (addTask / completeTask helpers).

claude.ai web?

Still no: custom connectors on claude.ai require a publicly reachable URL (Claude connects from Anthropic’s cloud, not your machine). To demo that, tunnel this server (e.g. ngrok http 3001) or deploy it — and note that real remote servers add OAuth, which this demo skips.

Notes

State is in-memory: restarting the process clears the list (persist to a file/SQLite as an exercise).
DELETE /mcp with a session id ends that session; GET /mcp opens the server→client notification stream.