Skip to main content
All posts

Introducing the trackagoat MCP server

Connect Claude, Codex, Cursor or another AI agent to your program, and ask it about your creators, goals and payouts.

· trackagoat

Right now, finding out who's behind on their goals means opening Creator Goals, picking the project and reading down the table. The same goes for checking what you owe, or adding a batch of new creators to a campaign.

With the trackagoat MCP server, you can ask an AI agent instead. Connect Claude, Codex, Cursor or another agent to https://www.trackagoat.com/api/mcp, and it can read and change your program the way you would in the app. This post covers what it can do, how to connect it, and what it can't touch.

What it is

MCP (Model Context Protocol) is the standard way AI agents connect to other software. The trackagoat MCP server has 125 tools: 63 that read your data and 62 that make changes. Each tool calls the same endpoint as the REST API, so your agent sees the same numbers you see in the app, with the same permissions and limits.

Here's what your agent can work with:

  • Creators and accounts. Add creators by their TikTok or Instagram handle, edit them, add notes, or stop tracking them.
  • Videos. Add a post by its URL, include or exclude videos, and review what's waiting in the Tracking Inbox.
  • Campaigns. Create campaigns, add creators, videos or other campaigns to them, and follow progress toward a view target.
  • Creator Goals. Set a goal like 3 posts a week, and see which creators are On Track, Behind Pace or At Risk.
  • Payouts. Set up payout structures, assign creators to them, see what each creator is owed, and record payments. It can also send payouts through Dots, once an org admin allows it.
  • Insights. Top videos and creators, posting consistency, campaign trends and Program Health.
  • Comment Insights. What people are saying in the comments, including anyone asking whether a post is an ad. Comment Insights is on paid plans.
  • The changelog. Ask what's new in trackagoat and the agent reads it from the get_changelog tool.

The full list of tools is on the MCP server page.

How to connect it

There are two ways to connect. Signing in is the easiest, because there's no API key to create or store. When your agent's client starts the sign-in, a trackagoat page opens. You pick the organization, choose View data and, if you want the agent to change things, Make changes. Then click Allow access.

Claude.ai or Claude Desktop

  1. Open Settings → Connectors (on claude.ai, Customize → Connectors) and choose Add custom connector.
  2. Name it trackagoat and enter https://www.trackagoat.com/api/mcp as the URL. Leave the OAuth client fields empty.
  3. Click Connect, sign in to trackagoat, pick your organization and click Allow access.
  4. Turn the connector on in a chat and ask about your creators.

Claude Code, Codex, Cursor and VS Code

In Claude Code, run this, then run /mcp, choose trackagoat and sign in:

claude mcp add --transport http trackagoat https://www.trackagoat.com/api/mcp

In Codex, run these two commands. The second one opens the sign-in:

codex mcp add trackagoat --url https://www.trackagoat.com/api/mcp
codex mcp login trackagoat

In Cursor and VS Code, add the URL to the editor's mcp.json file and sign in from its MCP settings. The exact config for each one is in Set up with a coding agent.

If a client can't sign in, use an API key instead. An org admin creates one in Org settings → API keys, and you keep it in an environment variable like TRACKAGOAT_API_KEY. The API key steps show how each client reads it.

Or let your agent set it up

The home page has a Set up trackagoat with your coding agent bar at the bottom. Copy the prompt and paste it into Claude Code, Codex or Cursor. Your agent walks you through signing up if you're new, adds the MCP server and checks the connection by listing your projects. You do the sign-ins yourself, and the prompt tells the agent never to type your password or ask for your API key in the chat.

What you can ask it

These are things you can type as is. Your agent picks the tools and reads the results back to you.

"Which creators are behind on their weekly goal?" The agent checks Creator Goals for the project and lists everyone who's Behind Pace or At Risk, with their target and how far along they are. For instance, a creator with 1 of 3 posts and 2 days left in the week.

"Add @handle to the Summer campaign." If the creator isn't tracked yet, the agent adds them first, which counts toward your plan's account limit. Then it adds them to the campaign. This needs Make changes.

"What do we owe creators for September?" The agent reads the Payouts tracker for September and gives you what each creator earned, what's been paid and what's still owed, largest first.

"Pay @handle what they're owed through Dots." The agent first checks that the creator is linked to a Dots payee and that the amount fits your limits. It should confirm the creator and the total with you before it sends anything. This only works after an org admin turns on Agent payments (more on that below).

"Which 10 videos got the most views in the last 30 days? Group them by creator." The agent ranks your tracked videos by views in that range and groups the top 10 by creator.

"Is anyone in the comments asking whether our posts are ads?" The agent pulls the comment flags for the last 7 days: comments asking whether a post is an ad, or using words like scam, most liked first. This needs a paid plan.

"How is the program doing this week? What should I look at first?" The agent reads Program Health and walks you through the alerts, like creators who stopped posting or accounts that went private.

"What's new in trackagoat?" The agent reads the changelog and summarizes recent changes, with links to each entry.

Permissions and limits

Your agent acts as you. A signed-in connection can only do what your role in the organization allows. If your role changes or you leave the organization, its access changes or stops at the same time. With View data only, a write tool is refused and nothing is changed.

One organization per connection. Each sign-in connection and each API key belongs to one organization. Your agent never sees another organization's data. To use a second organization, connect again and pick it.

Read and write are separate. API keys have a Read scope, which every key gets, and an optional Write scope. Sign-in has the same split as View data and Make changes, and the MCP server never uses anything beyond these two. Tools that delete, remove or void something are marked as such, so most clients ask you before running them.

Some things stay in the app. The MCP server can't create or revoke API keys, invite or remove members, or create or rename projects. It can't permanently delete conversions, and removing a video stops tracking it instead of deleting it. Staff tools for running trackagoat itself aren't exposed either.

Sending money is off by default. An org admin turns on Agent payments in Org settings → Payment providers, with a limit per payout and a limit per 24 hours. They start at $1,000 per payout and $5,000 per 24 hours, and you can change or clear them. Even then, only a connection or key that belongs to an org admin can send a payout, and one over a limit is refused with nothing sent. No API key can turn the switch on. See Paying creators with Dots.

Every tool call is one API request. Calls count toward your organization's daily request limit, which resets at midnight in the organization's time zone. get_changelog is the exception: it reads public release notes, so it doesn't count. Each key or connection can also make up to 60 calls a minute.

Every plan includes the MCP server and the API, including Free:

PlanPriceAPI requests a dayAPI keys
Free$02001
Starter$29/month1,0001
Ultra$399/month50,0005

Signing in doesn't use up one of your API keys. The rest of each plan's limits are on Plans and limits.

Read more

If you're not sure where to start, ask your agent what's new in trackagoat. That one's free.