MCP server

The Model Context Protocol server exposes your workspace to AI agents as structured tools. A todo filed by an agent is the todo you see in the app — same workspace, same real-time sync — and the app shows it was the agent’s work.

Tools

• Todos — list_todos, get_todo, save_todo (create or update: name, list, description, acceptance criteria, design decisions, notes, priority, assignee, due date, repeat rule, state, parent, labels), delete_todo, restore_todo
• Bulk — bulk_update_todos and bulk_archive_todos change up to 1000 todos in one call, selected by identifiers or by filter, with a dryRun to preview
• Relations — add_relation, remove_relation (blocks, blocked by, related, duplicate)
• Lists and labels — list_lists, get_list, save_list, list_labels, save_label, assign_label, remove_label
• Comments — list_comments, save_comment, delete_comment
• Questions — create_questionnaire, get_questionnaire, list_questionnaires, dismiss_questionnaire (questionnaires)
• Memory — remember, search_memories, delete_memory (memory)
• Workspace — list_workflow_states, list_members, workspace_health

save_todo warns when a new todo lacks a description or acceptance criteria, and when a todo is closed with unticked criteria — nudges that matter when agents are the ones filing.

Hosted endpoint

Nothing to install — point a Streamable HTTP client at the hosted server with an API key as a Bearer token:

https://api.createtodo.com/mcp
Authorization: Bearer <your-api-key>

For example, in Claude Code:

claude mcp add --transport http createtodo https://api.createtodo.com/mcp \
  --header "Authorization: Bearer $CREATETODO_API_KEY"

Clients that speak OAuth, such as claude.ai connectors, can add the same URL without a key: the server advertises its authorization endpoints and signs you in through the app.

Run it locally

Prefer a local process? The same server runs over stdio from npm — add it to any MCP client config:

{
  "mcpServers": {
    "createtodo": {
      "command": "npx",
      "args": ["-y", "@createtodo/mcp"],
      "env": { "API_KEY": "<your-api-key>" }
    }
  }
}

MCP or CLI?

The MCP server is the right fit for chat-centric clients that speak tool calls natively. For coding agents with shell access, the ct CLI is often cheaper: it returns compact JSON the agent can filter before it costs context, and it works offline. They share the same tool implementations — mix freely.