Trybe MCP vs API, and how we took Trybe to 37% of a 7-figure Meta account
· Paid Social

Most brands we talk to open a Trybe program, get 500 videos in 2 weeks, and launch about 20 of them. The other 480 sit in a queue. Creators lose interest and stop posting, so the brand concludes Trybe doesn’t work.
One account we run went from 0% to 37% of its Meta spend on Trybe content in 8 weeks. Total account spend rose almost 50%, to over 7 figures a month, while Trybe’s CPA held within 5% of the rest of the account. What moved those numbers was the automation we built around the creators.
We built it on Trybe’s API and on Data Druid, our in-house account intelligence platform. Claude connects to Data Druid through an MCP (Model Context Protocol, a connector that lets Claude read and act on a platform’s data). Below is what each one is for, the technical details if you want to build it yourself, and the full stack.
API vs MCP, in one line
Use the API for anything you want to happen every day without asking, and the MCP for the questions you’d otherwise send to an analyst.
The technical details, if you’re building it yourself
Trybe issues one API key per brand, scoped to that brand’s data only. For an agency running several clients, that made the API the natural base.
What we pull from it:
- Creators and their commission rules
- Every submission and its status
- Performance per creator for any date range, sortable by any metric
Lists return 100 rows per page, so a large program needs pagination. We sync everything nightly into a sheet or database and build on that copy, which is faster than querying live and keeps you clear of rate limits.
The stack
A launch queue (API)
Every approved video that isn’t live yet, pulled from Trybe and sorted oldest approval first, then ranked by internal rating, creator history and product. The media buyer opens it every morning and launches the top of the list in a couple of clicks. Our backlog went from 500 videos to a list we clear in a day.
A scaling SOP Claude runs every morning (Data Druid)
Our scaling rules are written the way you’d brief a new buyer. Each morning, Claude reads them through Data Druid, which holds Trybe, Meta and Triple Whale data in one place. It checks every ad set and writes a plan that sorts each one into scale, hold or step down. A buyer reviews the plan and makes the changes.
A leaderboard on real revenue (API)
Trybe creators joined to Meta spend and to Triple Whale or Northbeam revenue. Every platform’s own reporting flatters its own creators, so we rank on third-party numbers.
Every ad traced back to its video (API)
Each live ad links to the creator and the submission behind it. When an ad wins, you can see who made it and which line did the work. That line goes straight into next week’s brief.
A rebook list (API)
Creators whose ads beat account CPA, ranked by time since their last post. We check it every Monday.
Stale approval alerts (API + Slack)
Anything approved more than a few days ago that still isn’t live gets flagged in Slack, because approved content sitting in a queue is money you’ve already spent.
The prompts we run in Claude
Trybe on its own:
- “Show me everything approved that we haven’t launched yet.”
- “List the creators we should rebook this month, with the reason for each.”
- “Pull the 5 strongest hooks from every transcript for this product.”
- “Write next month’s brief from our top 10 ads.”
- “Flag any approved video that mentions a claim we can’t make.”
With Trybe, Meta and Triple Whale together in Data Druid:
- “Rank creators by new customers in Triple Whale, separate from Meta-attributed sales.”
- “Launch these 3, pause these 2. Show me the plan before you do it.”
What’s next on our list
- A hook library built from every transcript, ranked by CPA
- A compliance check that scans transcripts before launch
- A gap finder for angle and product combinations nobody has made yet
The rule we use
If we ask Claude the same question twice, we build the answer into a feature on the API.
It’s also why we lean on the API over Trybe’s own MCP. In Claude, one MCP connection is tied to one brand’s login, which means only one Trybe account can be connected at a time. That gets painful fast across several programs, so we keep the MCP for quick questions and run everything daily on the API.