The conversation is the surface
Status, proposals, and receipts appear inline, in plain language. It's where you review findings, edit a plan, and approve — but message bodies never store raw dataset rows.
Cernavo is a conversation you can read, backed by a run you can trust. The conversation shows status, proposals, and receipts; the canonical run, artifacts, and retention state live on the server. You stay in control at every step.
You read and act in the conversation. The truth of what happened lives on the server — so the receipt you see and the record we keep stay in agreement.
Status, proposals, and receipts appear inline, in plain language. It's where you review findings, edit a plan, and approve — but message bodies never store raw dataset rows.
The canonical run, its artifacts, and retention state live server-side. Proposal and event payloads don't store upload bytes — the workspace stays honest about what actually ran.
The workflow-state rail isn't decoration — it reflects a sequence the product enforces. Approve first, execute deterministically, then verify and hand you the evidence.
Nothing touches your dataset until you approve a specific plan. Your approval records exactly which tasks were authorized.
Approved tasks run as fixed transforms. The same input and the same plan always produce the same output — no silent or generative edits.
The result is checked against the approved plan, then handed to you with a change manifest and an approval record you can keep and audit.
The private beta is deliberately narrow. Knowing the edges is part of trusting the middle.
Request access to the private beta, or walk through the full run step by step.