Built for autonomous vetting
Multi-platform data, fake-follower detection, and brand-fit scoring — all behind one clean API your agent can call.
@levelsio · 587K followers
One endpoint, every creator
Vet creators across X, Instagram, TikTok, YouTube, and LinkedIn from a single API. We normalize the data so your agent doesn't have to learn each platform's quirks — pass a handle or link, get back a 0–100 score in about 30 seconds.
{
"handle": "levelsio",
"platform": "x",
"score": 78,
"tier": "A",
"subScores": { ... },
"redFlags": [ ... ],
"cpmRange": { low: 50, high: 125 }
}Drop-in for OpenClaw, Hermes, or Claude
Structured JSON, REST auth with API keys, no middleware. Pipe results into your agent of choice — or open the dashboard for a manual review. Same scoring model, same data shape, either way.
Not vanity metrics
Follower counts lie. We score reach, audience authenticity (with bot detection on a follower sample), niche focus, and conversion intent — then surface red flags so you don't pay for a misfit.
How it works
From handle to verdict in three calls — about 30 seconds end to end.
Drop a handle or link
Paste any creator from X, Instagram, TikTok, YouTube, or LinkedIn — or hit /api/score from your agent. We auto-detect the platform.
We pull and analyze
100 recent posts, a follower sample, and an LLM pass over both — bot detection, niche focus, brand mentions, and red flags.
Score, ready to act on
Structured JSON — a 0–100 score, four sub-scores, red flags, suggested CPM, and a shareable report. Drop it into OpenClaw, Hermes, or Claude.
Questions, answered
Everything worth knowing before you pipe creators through the API.
CollabPal is an autonomous vetting layer for creator sponsorships. Pass a handle or profile link from X, Instagram, TikTok, YouTube, or LinkedIn, and we pull engagement data, sample followers, run an LLM analysis, and return a structured 0–100 score plus a detailed report — built for AI agents (OpenClaw, Hermes, Claude) and the humans running them.
X today, with Instagram, TikTok, YouTube, and LinkedIn rolling out next. One endpoint normalizes the data so your agent doesn't have to learn each platform's quirks — same JSON shape no matter who you audit.
Four weighted sub-scores: Reach (40%) — log-scaled followers + median impressions + engagement rate. Audience (30%) — follower/following ratio + posting velocity + bot detection on a sampled follower set. Content (20%) — niche concentration of top topics. Conversion (10%) — CTA cadence + brand mentions + bio link. Composite rounds to 0–100 and maps to a tier S/A/B/C.
Yes — the Agent API plan gives you a REST endpoint (/api/score?handle=…) with API-key auth and 1,000 calls per month. The response is the same JSON the dashboard renders, so you can drop it straight into OpenClaw, Hermes, Claude, or any custom agent without translation.
Each audit hits the platform's API live, so the data is as fresh as the platform allows. Reports cache by creator ID for 24 hours — if anyone audits the same handle within that window, they get the cached report for free.
We can't audit private accounts — we need to read recent posts. Suspended and not-found accounts return a clear error and don't burn a credit. Locked accounts produce a low-confidence score with a red flag.
Because follower counts are gameable and don't predict ROI for brands. We weight engagement-per-impression, follower authenticity, and niche focus heavily — so a 5K-follower niche creator can outscore a 500K generalist with bought followers.