Skip to content

Any agent, full teammate

Everyone gives agents a seat. We give them a brief.

Connect Cursor, Claude Code, Codex, VS Code or any MCP client with a URL and a token. Agents read the brief, start the ticket, run a timer, link the pull request and prove every check, with the same rights as the person they work for.

Start freeSee plansEvery plan including Free
Cursor · agentMCP · newwork-projects
  • ●pm-item-getbrief, 4 checks, readiness 100
  • ●pm-item-startIn progress · branch chk-142-export-hours-csv
  • ●pm-timer-startvia AI agent
  • ●pm-check-tickpassed · PR #381
  • ●pm-comment-add@mira the separator is now configurable
  • ●pm-item-statusin_review · 1.4 h logged

Add to .cursor/mcp.json in your repository, or ~/.cursor/mcp.json for every project:

{
  "mcpServers": {
    "newwork-projects": {
      "url": "https://app.newworkprojects.com/api/mcp",
      "headers": {
        "Authorization": "Bearer nwmcp_…"
      }
    }
  }
}

Create the token in Workspace settings → Agents & MCP. It acts as you, in one workspace.

The app’s own components, rendered with example data.

§ 01

Connect by URL, or with one npx command

NEWWORK Projects speaks the Streamable HTTP transport of the Model Context Protocol. Paste the URL and a personal token into your agent, or run the newwork-projects-mcp package for clients that only start local servers.

  • Tokens are personal and belong to one workspace. Create and revoke them in Workspace settings → Agents & MCP.
  • Setup guides for Cursor, Claude Code, Codex and VS Code are in the docs.
§ 02

Over 30 tools that cover the whole loop

Agents do not just read tickets. They do the work around the code that people used to do by hand.

  • Find work: pm-my-work, pm-items-list, pm-item-get with the brief, checks, readiness and comments.
  • Start: pm-item-start moves the ticket to In progress and returns a branch name, a commit prefix and a warning when the brief is not ready.
  • Prove: pm-check-tick passes a check with evidence, pm-pr-link attaches the pull request.
  • Track: pm-timer-start and pm-timer-stop, or pm-time-log, so agent hours land on the timesheet.
  • Talk: pm-comment-add with mentions, notifications, watchers and attachments.
  • Plan: pm-item-upsert and pm-items-create-batch write new tickets as briefs, and pm-brief-check scores a draft before it is saved.
  • Read specs: pm-spec-get returns a spec with its sources, so the agent builds from the contract, not a summary of it.
  • Document: pm-docs-search and pm-doc-get read the product’s docs, and pm-doc-proposal-submit proposes updates that quote the diff (Business).
§ 03

The same rights as the person

An agent acts as the owner of its token: it sees the projects that person sees and changes only what they may change. Everything it does shows in the history with “via AI agent”, so you always know who, or what, moved a ticket. On Business, the audit log keeps every change with what it was before, for a year.

§ 04

Hand a ticket to a coding agent

Business

Assign to a coding agent in a ticket’s Code tab hands it to a Cursor cloud agent with the brief and the acceptance checks. It works on its own branch, ticks checks with evidence and opens a pull request named with the ticket key, so it links itself. When it finishes, the ticket moves to In review and you are told.

  • The agent gets a token of the person who started it, valid until the run ends and at most 24 hours.
  • The repositories assigned to the project come first; GitHub and GitLab are connected in Cursor.
§ 05

Never metered

MCP calls are not AI actions. They are not counted and not limited on any plan, including Free. Your agent runs on your model and your subscription; we do not sit in between or charge per call.

Questions

Which agents work?

Any client that supports MCP over Streamable HTTP or stdio: Cursor, Claude Code, OpenAI Codex, VS Code with Copilot, Windsurf, Zed and others. The docs have copy-paste setups for the common ones.

Does an agent need its own seat?

No. An agent works with a person’s token and acts as that person. Seats are counted per member, and agents are never billed.

Does our code go through NEWWORK Projects?

No. MCP exchanges ticket data only: briefs, checks, comments, time and links. Your code stays between you, your repository host and your agent.

Can we take an agent’s access away?

Revoke the token in Workspace settings → Agents & MCP and the next call fails. Someone removed from the workspace loses access through their tokens as well, because every call checks membership again.

Related

Give your next agent a brief.

Free for up to 10 people, with MCP for every agent. No credit card.

Start freeCheck a ticket