Todo4you now ships with a hosted Model Context Protocol (MCP) server. That means you can point Claude, Cursor, Zed, or any other MCP-aware assistant at your project and simply talk to it in natural language. No scripts, no copy-pasting from a chat window to your project tool - your assistant reads and writes tickets directly.
This post walks through what the integration can do now, how to set it up in a few minutes, and the kind of prompts that get the most out of it.
Why MCP?
If you are already spending half your day in an AI assistant, jumping to a browser tab just to file a bug or move a ticket breaks your flow. MCP closes that loop. Your assistant becomes the interface to your project:
- Draft a ticket from a half-formed thought without leaving your editor.
- Ask "what is still open in this sprint?" and get a real answer from live data.
- Move a ticket to "Review" the moment you push a commit, without touching the kanban.
- Start a timer on a ticket the moment you begin work, and stop it when you context-switch.
Because everything runs through a scoped token you generate yourself, you decide exactly which projects the assistant can see and whether it is allowed to write.
What the MCP server can do
The toolset grew a lot in this release. The server now exposes 37 tools covering the full ticket lifecycle - from reading and searching through to time tracking and checklist management.
Projects and metadata
list_projects,project_stats,project_statuses- discover what the token can see and summarise a project.list_project_members,list_tags- resolve user IDs for assignment, and look up tag IDs.
Tickets
list_tickets,search_tickets- browse or full-text search the project.get_ticket- pull full detail: description, comments, checklist, tags, assignees, blockers, attachments, and a time summary.create_ticket,update_ticket- create or edit title, description, type, priority, deadline, estimation, and tags.move_ticket,archive_ticket,delete_ticket- progress a ticket through statuses or take it off the board.assign_member,unassign_member- manage who is responsible.
Comments
list_comments,add_comment,update_comment,delete_comment- manage the conversation on a ticket.
Checklist and blockers
add_checklist_item,toggle_checklist_item,update_checklist_item,delete_checklist_item- break work into subtasks.add_blocker,remove_blocker- track ticket dependencies.
Tags
create_tag,update_tag,delete_tag- manage the tags of a project.
Time tracking
start_timer,stop_timer,pause_timer,resume_timer- run a live timer on a ticket.add_time_entry,update_time_entry,delete_time_entry- log or adjust time manually.list_time_entries,running_timer- read what has been logged and what is ticking right now.
Everything routes through the same /api/v1 REST endpoints the web app uses, so permissions, custom statuses, and notifications all behave exactly as they do in the UI.
Setting it up
- Go to MCP access settings and create a token. Pick read-only or read/write, choose which projects it can touch, and optionally set an expiry date.
- Copy the token once - it is hashed in the database and never shown again.
- Add the hosted server URL and your token to your MCP client. The settings page has copy-paste snippets for Claude Code, Claude Desktop, Cursor, Zed, VS Code with Continue, and any generic HTTP client.
- Start a new chat in your assistant - it will discover the tools automatically.
For a longer walkthrough and the security model, see the MCP feature page.
Example prompts
The fun part. Here are prompts that actually work well once your assistant is connected.
Triage and planning
"List all projects I have access to and tell me which one has the most open urgent tickets."
"In the
todo4youproject, how many tickets are sitting in Review right now? If it is more than five, suggest we do a review round."
"Search
webshopfor tickets mentioning 'VAT' and give me a one-line summary of each."
Creating and shaping work
"Create a bug in
apititled 'Invoices PDF missing VAT line' with high priority. In the description, explain that customers on the Dutch plan see a blank VAT row on invoices generated after April 10, and note that the reproduction case is in ticket API-214."
"Open TDL-120, set priority to urgent, assign it to Bob, and break it into checklist items: design, implement, write tests, update changelog."
"For TDL-88, add a blocker pointing to TDL-91, then leave a comment explaining why."
Keeping the project in sync while you code
"Move TDL-47 to 'review' and add a comment summarising what I changed: switched the session cookie to SameSite=Lax and added a regression test in
tests/unit/AuthTest.php."
"Start a timer on TDL-47 - I am picking it back up."
"How much time did I log on TDL-47 this sprint, and which entries were longer than an hour?"
Status updates for stand-ups
"Draft my stand-up: what changed on the
todo4youproject in the last 24 hours, who owns what is in progress, and which tickets look stuck."
A few tips
- Use slugs, not labels. Statuses are matched by slug (
in-progress,review,done). If your assistant gets confused, ask it to callproject_statusesfirst. - Assignments need user IDs. Ask for
list_project_membersonce at the start of a session so your assistant has the IDs ready. - Scope narrowly. Give each assistant its own token with only the projects it needs. Revoke and re-issue is a two-click flow.
- Read-only first. If you are still figuring out what to ask, start with a read-only token. You can always upgrade later.
- Trust, but verify. AI assistants occasionally hallucinate ticket IDs or priorities. The project board is the source of truth - glance at it after a batch of writes.
Ready to try it?
Head to /profile/mcp-access to generate your first token, or read the full MCP overview if you want more background first. If you build something interesting on top of it, let me know - info@todo4you.com.