How it works
Every account gets its own AI agent, running in the cloud on Omnara. It reads your posts, decides what to dig into, and writes the verdict. Here's everything that happens between your reply and ours.
You reply on 𝕏
Any reply works. Tag a friend to judge them instead.
We pull your posts
Your profile and last 200 posts, with views and likes.
list_postsyour best posts✓list_postsyour flops✓list_postsyour replies✓set_topics4 topics✓submit_reportscore + roast✓An agent investigates
A fresh agent on Omnara decides what to read, calls tools to dig in, then writes the verdict.
You get judged
We reply with your card. Your report has a chat with the same agent.
Why a cloud agent
Not a single prompt to a model. A real agent that lives on Omnara, uses tools, and keeps its memory.
Every judgement starts a fresh agent on Omnara. They run side by side in the cloud, not on our server.
The agent decides what to read: your best posts, your flops, your replies. It sorts your posts into topics before it scores anything.
The agent doesn't end when the report is done. Ask it something on your report page and it answers with the full context of your account.
How the score works
Five scores out of 10. Your overall score is their average.
The agent's call
Read from your posts and profile, the way a sharp friend who's great at 𝕏 would.
- Content
- Profile
- Originality
Computed from your numbers
The same posts always get the same score here, no matter which agent runs.
- Consistency
- Engagement
Build your own
The judge is one Omnara agent profile with three custom tools. Custom tools aren't run by Omnara: when the agent calls one, our server gets the call and sends back the answer through the Omnara API.
That's the whole pattern. Your app keeps the data and the side effects, and Omnara runs the agent.
model: provider_config: omnara-openrouter name: anthropic/claude-haiku-5.5 tools: list_posts: type: custom set_topics: type: custom submit_report: type: custom