Claude + Teamwork.com: What the Integration Can (and Can't) Do in 2026
Teamwork.com ships its own official MCP server in 2026, so connecting Claude is genuinely first-party — no third-party repo required. The server is hosted at https://mcp.ai.teamwork.com, it’s OAuth-secured and scoped to your account’s permissions, and Teamwork publishes a verified setup guide for Claude alongside ChatGPT, Copilot, and Cursor. If you’d rather build your own path, the Projects API v3 covers tasks, projects, milestones, and time entries directly.
Here’s each route, when to pick it, and what to use when you want Teamwork.com work that happens without you at the keyboard.
Route 1: The official Teamwork.com MCP server (remote connector)
This is the route to reach for first. Teamwork.com launched its official MCP server in 2026 and made it available to all paid customers. An admin enables MCP from Settings, then you add it to Claude as a custom connector.
In Claude Desktop, open settings, go to Connectors, choose Add custom connector, and enter the server URL:
https://mcp.ai.teamwork.comClaude opens an OAuth flow — you sign in to Teamwork.com and approve access. From then on Claude can read and act on only the projects, data, and actions your Teamwork.com account already has permission for. The toolset covers creating projects, assigning and updating tasks, adjusting schedules, logging time, and generating client-ready reports from a natural-language request. There’s no extra charge from Teamwork.com for the MCP server, though you’ll pay for whatever AI tool you connect.
Because it’s a remote/hosted connector (not a local process), this is the cleanest option for most people — nothing to install or keep running.
Route 2: Self-host the open-source MCP server
Teamwork.com also open-sourced the server as Teamwork/mcp on GitHub — written in Go, MIT-licensed, roughly two dozen stars but under active development with a fast release cadence (v1.24.x at the time of writing). It’s the same codebase behind the hosted endpoint, so self-hosting is for teams that want the server inside their own network or wired into a desktop app directly.
It ships two transports: an HTTP server for cloud/multi-client setups and a STDIO server for direct desktop or development use. Auth is via bearer token (TW_MCP_BEARER_TOKEN) or OAuth2, and there’s a read-only mode if you want Claude to look but not touch. It needs Go 1.26+ (a Dockerfile is included). For a stdio setup, point your Claude Desktop claude_desktop_config.json at the binary and pass your Teamwork bearer token as an environment variable.
Pick this only if you have a reason to run your own instance — otherwise the hosted connector in Route 1 is less work.
Route 3: The Teamwork.com Projects API v3 (build it yourself)
If you’re scripting with Claude in Claude Code or wiring Teamwork into your own agent, go straight to the API. The Projects API v3 gives programmatic access to projects, tasks, milestones, time entries, people, companies, tags, and reporting.
Two auth methods:
- API key (Basic auth) — the fastest. In Teamwork, click your avatar → Edit my details → the API tab → Show your token. Send it as the username in an HTTP Basic
Authorizationheader (over HTTPS only). - OAuth 2.0 (App Login Flow) — register your app on Teamwork’s Developer Portal for a client ID and secret, then make requests with a
Bearertoken. This is the right choice for anything multi-user or distributed.
Hand Claude the endpoint shapes and your key, and it can draft the requests to pull a project’s open tasks, create a task, or log a time entry. This is the most flexible route and the one that underpins scheduled automation — which is exactly where the MCP routes hit their ceiling.
One note: Teamwork.com has its own built-in assistant, TeamworkAI, and the MCP server is explicitly the foundation for its agentic features. So some of what you might reach to Claude for already lives inside the product.
The limits that actually matter
These assistants can run on a schedule now — ChatGPT Scheduled Tasks and Claude Cowork both do unattended work on a timer. What none of them do is fire on an event: nothing starts because a record changed, a message landed, or a deadline passed. The schedule is the trigger, and the gap between "every hour" and "the moment it happens" is where the work falls through.
- No event triggers. Nothing fires when a task is completed or a deadline slips — MCP tools run when you ask, or when a schedule you set comes around.
- Laptop-bound. A stdio/desktop MCP setup needs Claude Desktop open and running on your machine. Close the lid and the automation stops.
- The web app can’t reach a local server. claude.ai in the browser can use remote connectors like Teamwork’s hosted endpoint, but it can’t touch a locally-run stdio server or your Claude Code environment.
- Scoped to you, on demand. The connector is bound to one user’s permissions and only acts when prompted — great for pulling and updating, not for standing watch.
Claude + Teamwork.com is excellent for asking questions and making changes in the moment — “spin up this project, assign these tasks, log two hours.” It’s not built for work that has to happen the moment a task closes or a deadline slips.
If you want Teamwork.com work that runs on its own: Carly
Carly is an AI executive assistant that runs on triggers and schedules — no session to babysit, no laptop to keep open.
- When a task is marked complete, Carly emails the client a status update and logs the billable time against the project.
- Every Friday at 4pm, Carly pulls the week’s completed tasks and time entries and sends the account owner a client-ready summary.
- When a milestone due date passes with open tasks, Carly pings the project lead in Slack and flags the at-risk deliverable.
- No-code setup. There’s nothing to configure by hand — Carly interviews you about the workflow in plain English and builds it.
- Actually sends — Carly drafts and sends real email through Gmail and Outlook, not just draft suggestions.
- Connects to anything — 200+ native integrations plus any tool with an API via your own key. Carly reaches Teamwork.com through your own Teamwork.com API key (paste it on the integrations page) — one of 200+ native integrations plus anything with an API. There’s no one-click native Teamwork connector; the BYO-key path is how it wires in.
AI agents start at $35/month, and steps in a workflow that don’t use AI run free and unlimited.
Claude vs Carly
| Claude + Teamwork.com | Carly | |
|---|---|---|
| Read/search Teamwork projects & tasks | Yes (MCP or API) | Yes |
| Create/update tasks in a session | Yes | Yes |
| Acts on triggers (task done, milestone slips) | No | Yes |
| Runs on a schedule | No | Yes |
| Acts the moment something happens | No | Yes |
| Sends email (Gmail / Outlook) | No | Yes |
| Setup | Add connector or wire the API | Interview, no code |
| Pricing | Pro $20 / Max $100–$200 | AI agents from $35/mo |
Claude is the best way to work with Teamwork.com in the moment; Carly is the way to make Teamwork.com work happen on its own.
Frequently Asked Questions
Does Claude work with Teamwork.com?
Yes. Teamwork.com ships an official, OAuth-secured MCP server at https://mcp.ai.teamwork.com that connects to Claude Desktop as a custom connector, available to all paid customers. You can also wire Claude to the Projects API v3 directly.
Is there an official Teamwork.com MCP server?
Yes — it’s first-party. Teamwork.com hosts it at https://mcp.ai.teamwork.com and also open-sourced the code as Teamwork/mcp on GitHub (Go, MIT-licensed) so you can self-host. It covers tasks, projects, timers, and reporting.
Can claude.ai on the web access Teamwork.com?
Yes, through the hosted remote connector, since that’s a URL claude.ai can reach. What it can’t reach is a locally-run stdio MCP server or a Claude Code setup on your machine.
Can Claude log time or update tasks in Teamwork.com automatically?
Not on its own. MCP tools don’t fire on events — there’s no “when a task closes, log the time” trigger. For that, use an automation platform like Carly.
More: Claude connectors explained · Claude + Asana · Claude + ClickUp · Best Smartsheet integrations · Can Claude send emails?
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
