Documentation
Analysis-only mode
Hard product boundary: observe and coach — never auto-send messages.
v1 is analysis-onlyChatalyze does not send messages to fans. Chatters continue to work in OnlyFans (or their CRM). We analyze traffic that arrives via OnlyFansAPI.
Why this exists
- Platform ToS and ban risk around automation
- Managers need trust: AI coaches, it does not impersonate at scale
- Cleaner legal/ops story for agencies
What is allowed
| Capability | Status |
|---|---|
| Ingest webhooks / backfill | Live |
| Metrics, SLA, priority inbox | Live |
| Alerts & recommendations (rules) | Shipped |
| Structured AI labels (optional HTTP model) | Shipped |
| Floor wallboard & scorecards | Shipped |
| Auto-send / bot replies | Out of scope |
Engineering guarantees
- Webhook handler never awaits an LLM (
processIngestPayloadis free of intelligence calls — regression testtests/api/webhook-no-intelligence.test.ts). - Intelligence runs via
POST /api/admin/intelligence/run(or future async queue). - ADR 006: rules before models · ADR 007: analysis-only v1.