Compliance Agents
ZebraTruth uses a pipeline of specialized AI agents, each focused on a specific compliance domain. Agents are added regularly — always useGET /agents to discover available agents dynamically.
Current Agents
| Agent | Phase | Input | What It Checks |
|---|---|---|---|
| Jurisdiction Detection | 1 | Text | Detects applicable laws (FTC, UCPD, ASA, etc.) |
| Advertising Law | 2 | Text | Claim substantiation, endorsement rules, targeting |
| Platform Policy | 2 | Text | YouTube, Instagram, Facebook, TikTok, LinkedIn rules |
| Metadata & Labeling | 3 | Text | C2PA Content Credentials, AI disclosure labels |
| Rights Clearance (Text) | 3 | Text | Celebrity references, trademarks, copyrighted characters |
| Rights Clearance (Image) | 3 | Image | Celebrity faces, logos, watermarks via computer vision |
| Compliance Score | 4 | All | Deterministic scoring from all checks |
Pipeline Phases
In full mode, agents run in 4 sequential phases with content enrichment:- Phase 1 — Jurisdiction agent detects applicable laws
- Phase 2 — Advertising Law + Platform agents (parallel), enriched with Phase 1 findings
- Phase 3 — Metadata + Rights agents (parallel), enriched with Phase 1+2 findings
- Phase 4 — Score agent aggregates all checks
3-Model Sub-Agent Pipeline
Each text agent internally runs a 3-stage pipeline:| Stage | Model | Role | Cacheable |
|---|---|---|---|
| A | Perplexity Pro | Legal research & citations | Yes (30-day TTL) |
| B | GPT-5.2 | Content-specific risk analysis | No |
| C | claude-opus-4-6 | Structured compliance writing | No |
Invoking a Single Agent
Call any agent directly viaPOST /agents/{agentId}. Costs ~8-12 credits vs ~47 for a full check.
See Individual Agents guide for details.