Exporting your data

The local JSON export bundle, what's included, where it goes, and how to use it for backup or migration.

Cerevisor's data is local-first. To back it up, migrate to a new machine, or just see exactly what's stored, use the export bundle.

The export button

Open the Memory view → Consent tab. Click Export everything as JSON.

You'll be prompted for a save location. Cerevisor writes a single JSON file.

What's in the bundle

The bundle (bundleVersion: 2) contains:

  • consent: your current consent flags + the full consent history (append-only audit log).
  • inventory: every record in the metadata table, with classification, provenance, and timestamps.
  • documents: the byte-for-byte content of every memory document (profile.md, profile.json, harness/self.md, harness/self.json, harness/corrections.ndjson, meta-cognition/observations.md, meta-cognition/harness-self-portrait.md, entries/MEMORY.md, etc.).
  • quality: per-record reinforcement count + maturity.
  • facets: your structured user facets (seniority, country, industries, etc.).
  • workflowShapes: the last 200 workflow runs' structural metadata (agent count, providers used, costs).
  • outcomeAlignments: outcome alignment scores per session.
  • bundles: memory bundles you've created (Phase 3 marketplace primitive; usually empty in v1).

The bundle is intentionally complete. Nothing about your Cerevisor memory state is missing.

What's NOT in the bundle

  • API keys (those live in the OS keychain, not the memory subsystem).
  • Workflow files (.cerevisor / .cerevisor-world: back those up separately).
  • Skills (those live in ~/.claude/skills/ and ~/.cerevisor/skills/: back those folders).
  • Audit logs (~/.cerevisor/audit-logs/: append-only NDJSON, back up the folder).

For a complete machine backup, copy the whole ~/.cerevisor/ folder. The memory bundle is a focused export of just the memory subsystem, useful for sharing with another person or auditing.

Use cases

Backup

Run the export monthly. Keep the file in your usual backup location (cloud storage, external drive). If your machine dies or you need to roll back, you can restore.

Migration to a new machine

  1. Export on the old machine.
  2. Install Cerevisor on the new machine.
  3. (Manual step until v2's import flow lands:) place the bundle at ~/.cerevisor/memory/import.json.
  4. Restart Cerevisor; it'll prompt to import.

Today, the cleanest migration is just copying ~/.cerevisor/ wholesale. The export bundle's import path is a Phase 2 feature.

Audit

Open the JSON in any text editor. Search for "containsThirdPartyPii", "consentScope", or any specific term. Confirm Cerevisor is storing what you think it is, and nothing else.

Sharing with auditors / compliance

The bundle is structured enough that an auditor can verify what data Cerevisor stores about a user without needing direct machine access.

Privacy

The bundle is never sent anywhere by Cerevisor. The export is purely local: you click the button, Cerevisor writes a file to disk. No telemetry. No background upload. If the file leaves your machine, you put it there.

File size

For a heavy user (months of daily use, hundreds of runs), the bundle is typically 1–10 MB. Most of the size is the workflow-shape rollup and the run logs: the rest is small.

Bundle versioning

bundleVersion in the JSON declares the schema. v1.0.x ships bundleVersion: 2. Future versions will add fields. Importers that only understand v1 will read v2 bundles successfully (additive changes only); they just won't see the new fields.

Other backup paths

If you don't want to use the export bundle:

Want Do
Full backup Copy ~/.cerevisor/ to your backup destination.
Just memory Copy ~/.cerevisor/memory/.
Just skills Copy ~/.cerevisor/skills/ and ~/.claude/skills/.
Just audit logs Copy ~/.cerevisor/audit-logs/.
Just workflows Find your .cerevisor files (wherever you saved them).

Back to docs