MCP server

A project management MCP server with 43 tools.

Planner.day's built-in MCP server gives Claude, ChatGPT or any MCP client 43 tools over streamable HTTP: create and update tasks, run sprints, read dashboards and post in chat. Every call runs the same permission checks as the web interface, bounded by a sign-in or token you scope yourself.

In ChatGPT or Claude.ai

Add a custom connector pointing at the endpoint and sign in. You approve the permissions and the workspaces on a Planner.day screen, and can disconnect from Settings at any time.

No token to create, copy or rotate.

In an editor, CLI or automation

Claude Code, Cursor, VS Code, Windsurf, Zed, JetBrains and n8n take the endpoint plus a bearer token you scope yourself.

One command
claude mcp add --transport http planner \
  https://api.planner.day/api/mcp \
  --header "Authorization: Bearer tvmcp_…"

Endpoint https://api.planner.day/api/mcp · streamable HTTP · 120 requests/minute per token

The whole catalogue

The assistant discovers these itself, filtered to what your token actually permits, so it never offers an action it would be refused.

Tasks

16
  • list_tasks Every task in a list, with its statuses.
  • get_task One task in full: assignees, tags, fields, dates.
  • list_task_attachments Files attached to a task.
  • list_task_designs The Figma designs attached to a task.
  • get_attachment_download_url A signed link to fetch one attachment.
  • create_task New task in a list. Type defaults to the list’s own default.
  • update_task Patch title, description, priority, dates.
  • change_task_status Move a task to another status.
  • change_assignees Add or remove people.
  • change_task_tags Add or remove tags.
  • set_task_estimate Set estimated hours.
  • create_subtask Nest work, up to five levels.
  • delete_task Soft-delete. Requires its own scope.
  • list_statuses The status columns of a list.
  • create_status Add a status column.
  • get_my_assigned_tasks What the token’s user owns right now.

Projects & people

16
  • whoami Identity and effective authority of the calling token.
  • list_workspaces Workspaces this token may touch.
  • get_workspace_tree Spaces, folders and lists, nested.
  • create_space New space.
  • create_list New list, seeded with three statuses.
  • search_tasks Full-text-ish search across a workspace.
  • create_tag New workspace tag.
  • list_members Who is in the workspace.
  • invite_member Invite someone.
  • change_member_role Change a member’s role.
  • get_workspace_stats Totals, completed, overdue, by status and priority.
  • get_metric One dashboard metric, computed server-side.
  • get_person_analytics Per-person breakdown.
  • get_team_availability Who has capacity.
  • list_dashboards Available dashboards.
  • get_dashboard One dashboard with its widgets.

Collaboration

3
  • get_comments Comment thread on a task.
  • add_comment Post a comment.
  • get_activity The audit trail of a task.

Sprints

2
  • list_sprints The sprints in a sprint folder, in order, with each one’s state.
  • get_sprint One sprint’s progress: dates, goal, committed work and what is done.

Chat

6
  • list_chat_channels Channels you can see, with unread counts.
  • get_chat_messages Read a channel or one thread, page by page.
  • search_chat Full-text search across the chat you can see.
  • get_chat_decisions What the team marked as decided, newest first.
  • post_chat_message Post as you, labelled "via AI agent".
  • chat_message_to_task Turn a message into a task, linked back to it.

A token is the whole authority

Membership roles are not consulted for MCP. What a token can do is exactly its scopes intersected with its workspaces. Nothing widens that at runtime.

Same pipeline as the UI

Every tool call goes through the same handlers, permission checks and validation as a click in the app, and writes the same activity record. There is no faster, looser API behind the assistant.

Bounded twice

24 scopes and an explicit workspace list. The effective authority is the intersection, a token scoped to read tasks in one workspace can do exactly that.

No administrative ceiling to breach

No token can perform a super-admin action, including minting more tokens. That is enforced in the authorization pipeline, not left to configuration.

Revocable, expiring, observed

Every token has an expiry between 1 and 365 days, can be revoked instantly, and records when it was last used. Secrets are stored hashed and shown once.

Four presets, because scope lists invite over-granting

read_only_analyst

Reads everything it is given. Changes nothing.

Reporting, standups, weekly summaries.

task_bot

Creates and updates tasks. Cannot touch members or structure.

The assistant that files what you dictate.

pm_bot

The above plus dashboards and analytics.

A copilot that plans as well as records.

full_access

Everything the token’s workspaces allow.

Trusted internal automation.

MCP questions

Does Planner.day have an MCP server?

Yes. Planner.day has a built-in MCP server with 43 tools, so Claude, ChatGPT or any MCP client can create tasks, move work, read sprints and dashboards, and read or post in chat. Every call runs through the same permission checks as the person who connected it.

Which AI assistants can connect to Planner.day?

ChatGPT and Claude.ai connect as custom connectors: add the endpoint, sign in, and approve the permissions and workspaces on a Planner.day screen. No token is involved. Claude Code, Claude Desktop, Cursor, VS Code, Windsurf, Zed, JetBrains, n8n and anything else speaking Model Context Protocol over streamable HTTP connect with a scoped bearer token. The endpoint is a standard MCP transport, so nothing is specific to one vendor.

Can an AI assistant delete my data through MCP?

Only if you grant it. Deletion sits behind its own scope, separate from write access, so a token that can create and update tasks cannot delete them. Deletes are soft, so the row is retained and recoverable.

What stops an MCP token from being used outside its workspace?

Every token carries an explicit list of allowed workspaces. The authorization pipeline resolves the workspace of each request and rejects it if the workspace is not on that list, before any handler runs.

Is there a rate limit on the MCP endpoint?

Yes, 120 requests per minute per token. Exceeding it returns HTTP 429. The limit is per token rather than per user, so one noisy automation cannot starve another.

Stop transcribing decisions into a tool.

The assistant that heard the decision can record it. Free for up to 5 people, and the first month on Basic or Pro is free.