Handoff Record

Gigabrain's launch product surface is the Handoff Record (the verb is gigabrainctl handoff): the governance/audit artifact of the broader memory trust control plane. It is a local-first report that shows what your AI agents remember, where those memories came from, what needs review, and what can be safely handed to another host.

The Handoff Record is a report + safe handoff pack, not a portable re-importable bundle and not another hidden RAG store. It is the audit/trust layer above native memories:

• source inventory by host
• exact duplicate groups with retained provenance
• contradiction review items from the world model
• stale or expired memories
• provenance gaps
• secret-like memory risks with redacted previews
• a readiness verdict that separates launch blockers from cleanup work
• Markdown handoff briefs for AGENTS.md, CLAUDE.md, ChatGPT, Claude.ai, Gemini, and Microsoft Copilot, with secret-risk rows omitted entirely

Gigabrain ships the Handoff Record as a governance/audit + safe-handoff layer. Integrity-checked export/import bundles are available for deliberate transfer, but they are not encrypted. Vault-grade features such as encrypted bundles, signed manifests, and a complete review UI remain future work, not shipped today.

Note: the passport verb is a deprecated alias of handoff and still works (it prints a deprecation note). Use handoff in new scripts.

Quickstart

npm install @legendaryvibecoder/gigabrain
npx gigabrain-codex-setup --project-root /path/to/repo
npx gigabrain-claude-setup --project-root /path/to/repo
npx gigabrainctl sync-hosts --config ~/.gigabrain/config.json --host codex,claude_code,cursor,windsurf
npx gigabrainctl handoff --config ~/.gigabrain/config.json --output-dir ./gigabrain-memory-audit

Open ./gigabrain-memory-audit/memory-audit.md or ./gigabrain-memory-audit/memory-audit.html. Handoff Records are written under ./gigabrain-memory-audit/handoff-records/.

Manual Cloud Imports

Closed cloud products are explicit manual flows. Gigabrain does not scrape ChatGPT, Claude.ai, Gemini, or Microsoft Copilot memory, and it does not claim bidirectional sync with those systems.

npx gigabrainctl sync-hosts --config ~/.gigabrain/config.json \
  --manual-import ./chatgpt-memory-export.md \
  --manual-source-host chatgpt_manual \
  --scope profile:user

npx gigabrainctl handoff --config ~/.gigabrain/config.json \
  --scope profile:user \
  --output-dir ./gigabrain-memory-audit

Manual imports are tagged as manual_import with bidirectional_disallowed sync policy.

CLI

npx gigabrainctl handoff \
  --config ~/.gigabrain/config.json \
  --scope profile:user \
  --stale-days 180 \
  --output-dir ./gigabrain-memory-audit

Useful flags:

Flag	Purpose
--output-dir <path>	Directory for memory-audit.md, memory-audit.html, memory-audit.json, and handoff-records/
--format all|markdown|html|json|handoffs	Limit written artifacts
--scope <scope>	Limit report and handoff memories to one scope
--limit <n>	Max rows per audit section
--handoff-limit <n>	Max memories per Handoff Record
--stale-days <n>	Mark memories stale after this many days without update
--host <list>	Optional source discovery filter
--skip-handoffs	Write only the Memory Audit report files

Report Sections

Section	What it checks
Source Inventory	Synced source paths, host counts, sync policy, and freshness
Host Readiness	Local sources detected, last sync status, manual-only hosts, and bridge hosts
Readiness Verdict	A launch score, blocker list, and next actions based on the Memory Audit findings
Dedupe Audit	Active exact duplicate groups by normalized fingerprint
Contradiction Audit	Open contradiction-review loops from the world model
Stale Memory Audit	Memories past valid_until or not updated within --stale-days
Provenance Audit	Memories with registry-only or missing source provenance
Secret Risk Audit	Secret-like content using redacted previews only

Handoff Records never include secret-risk rows, not even with redacted text. If the Memory Audit finds credentials, tokens, or API-key-shaped memory, the brief records how many rows were omitted and points you back to the Secret Risk Audit.

Demo

Run the local launch demo script from the repo root:

npm run demo:handoff

The script creates a temporary standalone app, seeds synthetic Codex and Claude Code memory files, syncs them into Gigabrain, then writes a Memory Audit report and Handoff Records. It is safe to run because it uses a temporary HOME and CODEX_HOME.