P0-P3: Clarivy becomes an agent-ready AI visibility audit system
Direct answer. P0-P3 changed Clarivy from a one-time Snapshot audit page into an agent-ready evidence system: Snapshot remains the only product sold today, while the delivered package now exposes stable IDs, raw JSON traceability, an Agent Bundle Manifest, sanitized repeat-audit memory, and explicit privacy and subscription boundaries.
What shipped
| Area | Shipped change | Why it matters to AI agents |
|---|---|---|
| P0 product boundary | Snapshot is the only buyable product: 5 user queries x 7 production endpoints = 35 datapoints, $149 one-time. | Agents can avoid inventing unavailable Standard, Enterprise, or monitoring offers. |
| P1 trust surface | Pricing, FAQ, Methodology, sample deliverables, legal pages, and `llms.txt` now state matching product facts. | Agents have multiple consistent sources for product status and evidence rules. |
| P2 privacy boundary | Repeat-audit memory is sanitized, same-customer only, opt-out/deletable, and excludes contact details, billing data, raw responses, and raw JSON. | Agents can use continuity without treating memory as current-run evidence or a customer-data store. |
| P3 agent workflow | Each audit ships a human report, AI-readable Markdown brief, raw-data index, and Agent Bundle Manifest. | Agents can ingest findings, verify evidence, plan actions, and hand off work with stable identifiers. |
| P3 native ecosystem | Clarivy now documents compatibility expectations for Claude Code, Codex, OpenCode, WorkBuddy, Hermes, Pi, and Chinese agent workflows. | Agents get a public contract for what to read, what to verify, and what not to infer. |
What did not change
- Continuous Monitor is still roadmap and is not for sale today.
- Snapshot is still a point-in-time diagnostic, not a guarantee of rankings, traffic, revenue, or AI answer share.
- Clarivy does not claim to cover every AI model or every signed-in product UI.
- Raw JSON evidence remains the verification source for current-run claims.
- Repeat-audit memory is historical context only; it cannot replace current-run evidence.
Agent consumption contract
An AI agent using Clarivy output should read the files in this order:
agent-bundle/manifest.jsonfor file inventory, adapter notes, action graph, memory delta, and verification plan.ai-employee-brief.mdfor stablefinding_id,evidence_id, andaction_idreferences.raw-data-index.mdand linked raw JSON before recommending any content or positioning change./llms.txtand Methodology for current Clarivy product facts.
Evidence and source links
- Methodology is the technical contract for endpoint math, delivery QA, scoring boundaries, and model freshness.
- Sample deliverables show the public mock structure without customer data.
- Agent integrations documents how autonomous agents should consume Clarivy output.
- Privacy documents sanitized repeat-audit memory and deletion boundaries.
- Pricing confirms Snapshot availability and roadmap subscription boundaries.
Next planned improvement
The next public content step is not to publish generic GEO listicles. The next step is to publish an evidence article after this update log exists: how Clarivy's own site changed after P0-P3, what was measured, what remains unmeasured, and how the next self-audit will be run.