Your projects, inside your AI tools
A native MCP server with 31 tools. Claude Code, Cursor, and ChatGPT read and update real projects, bugs, and time — as you, and only as you.
claude mcp add --transport http worksynk https://api-v2.worksynk.app/mcp/worksynkEvery host, one URL.
No client ID or secret for hosts that register themselves — sign in once in the browser.
Claude Code
Add the server from your terminal, then run /mcp inside Claude Code to sign in.
claude mcp add --transport http worksynk https://api-v2.worksynk.app/mcp/worksynkClaude.ai & Claude Desktop
Settings → Connectors → Add custom connector, paste the URL, and approve the Worksynk screen.
https://api-v2.worksynk.app/mcp/worksynk
Cursor
Add Worksynk as a remote server; Cursor opens the browser to authorize on first use.
{
"mcpServers": {
"worksynk": {
"url": "https://api-v2.worksynk.app/mcp/worksynk"
}
}
}
VS Code
Add it as an HTTP server. The same config works in Windsurf and other VS Code–based editors.
{
"servers": {
"worksynk": {
"type": "http",
"url": "https://api-v2.worksynk.app/mcp/worksynk"
}
}
}
ChatGPT
Settings → Connectors → add a custom MCP server (availability depends on your ChatGPT plan).
https://api-v2.worksynk.app/mcp/worksynkServer URL: https://api-v2.worksynk.app/mcp/worksynk · Any other MCP-compatible host works too.
From bug report to fix in Claude Code.
Pick up a bug, fix it in your repo, and update the board — without leaving the terminal.
- 01
“What bugs are assigned to me in the Atlas project?”
Claude lists your open bugs, with codes, priorities, and due dates.
search_items - 02
“Read ATL-142 and its comments, then find the cause in this repo.”
It reads the report and the discussion, then works in your codebase.
list_work_item_comments - 03
“Move ATL-142 to In Review and log 45 minutes.”
With write access granted, the board and timesheet stay current without leaving the terminal.
update_task · log_time
31 tools, named as the host sees them.
Read tools need mcp:read. Anything that changes work needs mcp:write.
Find and read work
mcp:read- search_all
- Find tasks, projects, and clients by name, code, or title in one call.
- search_items
- Filter work items by title, code, state, project, or assignee.
- list_work_item_comments
- Read the comment thread on a work item — the why behind a decision.
- blocker_radar
- Items that need attention today: overdue, blocked, stale, or waiting on review.
- list_projects
- Projects you can see, with their delivery status.
- list_sprints
- A project's sprints, with status and story points.
- list_milestones
- The roadmap for a project, with dates and completion.
- list_meetings
- Upcoming meetings in the workspace.
- list_notes
- Your notes and the ones shared with you.
- list_workspace_members
- Who is in the workspace, to resolve an assignee.
- active_timers
- Who has a timer running right now, and for how long.
Delivery insight
mcp:read- project_status
- Item counts by state, overdue count, and retainer budget for one project.
- delivery_confidence
- A 0–100 confidence score for a project, with the factors behind it.
- explain_delay
- Root causes for why a work item is late or at risk.
- work_graph
- Tasks, dependencies, and health for a project as a graph.
- client_progress
- Progress across every project linked to one client.
- client_project_status
- A client-safe status: progress, recent completions, approvals.
- executive_report
- An AI-narrated health, budget, and risk report across projects.
Create and update
mcp:write- create_task
- Create a task, bug, feature, or story in a project.
- update_task
- Change state, priority, assignee, due date, title, or parent.
- log_time
- Log finished work time against a work item.
- start_timer
- Start a live timer on a work item.
- create_project
- Create a project, then shape it in the app.
- update_project
- Name, status, dates, or monthly hours cap.
- create_sprint
- Plan a sprint on a project.
- update_sprint
- Name, goal, dates, or milestone.
- create_milestone
- Add a milestone to a project's roadmap.
- update_milestone
- Name, description, dates, or status.
- create_meeting
- Schedule on Zoom, Google Meet, Teams, or a manual link.
- create_note
- A personal note, optionally on a project.
- update_note
- Edit, pin, complete, or set a reminder.
It acts as you, never more.
Built so you can connect an agent and still pass a security review.
OAuth 2.1 with PKCE
Hosts sign in through Worksynk. Hosts that support dynamic client registration set themselves up.
Read and write are separate scopes
Grant mcp:read alone and a host can look but never change anything.
Your permissions, exactly
A connection acts as the person who authorized it and can never reach a project they cannot.
The vault is out of reach
Stored credentials are never exposed to MCP, whatever scopes a connection holds.
Every call is logged
Tool calls are recorded against the workspace, so there is a trail of what an assistant did.
Revocable at any time
Owners and admins see every connection reaching the workspace and can revoke one instantly.
More on approvals and the AI assistant in AI agents & MCP and Security.
Common questions.
What is the Worksynk MCP server?
It is a Model Context Protocol server built into Worksynk. Any MCP-compatible AI host — Claude Code, Claude, ChatGPT, Cursor, VS Code — can connect to it and read or update your projects, tasks, sprints, milestones, notes, and time, using your own permissions.
How do I connect Claude Code to Worksynk?
Run "claude mcp add --transport http worksynk https://api-v2.worksynk.app/mcp/worksynk" in your terminal, then type /mcp inside Claude Code and sign in to Worksynk in the browser. No client ID or secret is needed — Claude Code registers itself.
Can Claude Code fix bugs tracked in Worksynk?
Yes. Claude Code can search the bugs assigned to you, read each one with its comment thread, fix the code in your local repository, and — if you granted write access — move the work item along and log the time spent.
Does the MCP server change data without asking?
A write tool runs as soon as the host calls it; there is no extra confirmation step over MCP. That is why read and write are separate scopes. Grant write access only to a host that needs it.
How do I turn the MCP server on?
MCP Server is a module. A workspace owner switches it on under Settings → Modules — the pricing page lists which plans include it.
Can an AI host see my stored credentials?
No. Vault contents are never reachable by MCP or any other integration, regardless of the scopes a connection holds.
/ Get started
Stop finding out about risk in the retro.
Score every project and catch scope creep early, without another status meeting.
- Free, full platform
- 90 days
- Nothing up front
- No card
- Guided walkthrough
- 20 min