Our dream agent wasn’t another chat panel assistant that answers questions already covered in the help docs. We wanted one that could do real Trust & Safety work. So we built the Musubi Agent to work directly with your policies, datasets, evaluation runs, and decision history. It can investigate a problem, test a possible solution, pull the examples behind the results, and propose a targeted change for you to review.
The Musubi Agent doesn’t replace the judgment behind good Trust & Safety work. It makes that judgment easier to apply. For experienced teams, that means more capacity to investigate, test, and improve. For teams still building their T&S function, the Agent brings Musubi’s playbooks into their daily work, while our human experts remain available for hands-on support whenever they need it.
At Stocktwits, Noah Loffredo uses the Musubi Agent both in the app and through its MCP extension, which brings the Agent into his coding workflow. In one day, he completed three policy iterations that would previously have required a much slower manual process.
“I actually did three iterations of the main policy very quickly, and got the accuracy up significantly in the first day that I was using it because I could rapidly iterate a lot quicker than it would have been manually.”
- Noah Loffredo, Senior Product Manager, Stocktwits
What using the Agent looks like in practice
Upload a few false-positive examples and ask, “How can we improve this policy?”
The Agent can then:
- Inspect recent decisions across recent production data
- Identify a recurring error pattern
- Ask for permission, then A/B test different policy changes against an appropriate dataset
- Check whether the changes improve results without creating new errors
- Give you a final recommendation, with data showing exactly how metrics have improved
That’s just one workflow. Here are some of the other ways the Agent can help:
- Understand what’s happening on your platform: Analyzes policies, datasets, evaluations, and production decisions to surface patterns and themes
- Evaluate performance: Runs evaluations and explains the errors behind the metrics
- Communicate the results: Produces executive-ready summaries, examples, charts, reports, and exportable data
- Design a workflow: Helps you choose between models, tailor policies for each, and design how content moves through an escalation path
When good T&S work becomes easier, teams do more of it
Early users of the Musubi Agent describe a change in the work itself: tedious tasks became easier, half-hour jobs took two minutes, and they can now act on ideas that previously weren’t worth the overhead.
Building golden datasets “has always been a pain,” says Noah, but now he adds to them every few days through the Musubi Agent. It samples decision history into a CSV, he validates the labels, and that becomes the baseline for the next round of iterations. “To do that on a weekly or daily basis is definitely something I would have never done before.”
Not a Musubi customer yet? Book a demo and bring your hardest policy. We’ll show you what the Agent can do with it.
And hey, if you already use Musubi, the Agent is included at no extra cost.
Wondering if Musubi could help?
We love to talk about the basics, the gnarly edge cases, and everything in between.