If GitHub is a thinking space for code review through commits and pull requests, glhub is a thinking space for evolution review through generations and evolution documents. Capture what AI agents change, why, and what carries forward.
Every generation gets a before / transition / after / next document with score deltas, gains, losses, retrospective rules, and bug fixes — not raw JSON.
Attach any GitHub, GitLab, Codeberg, Forgejo, or self-hosted Gitea URL. Provider auto-inferred. The lineage stays portable when forges change.
Local YAML-on-disk store, init / new / show / lineage / push commands. glctl login uses GitHub OAuth via loopback flow.
Comments and edits are saved as child generations, never overwriting evolution memory. The audit trail is the product.
branch?: string on every record. Use git worktrees to run parallel agent experiments and compare lineages side-by-side.
Lineage primitives are Apache-2.0 in the public repo. Multi-user RBAC, SSO, audit-grade compliance live in Cloud / Enterprise — clearly separated, no source-available trickery.
Every push is checked against the spec. Bad payloads get HTTP 422 with a human-readable errors[] list — the lineage stays clean.
Each metrics.score must be a finite number greater than 0 and at most 1. Score 0 is rejected — it indicates no actual evaluation. For failed runs, set success: false instead.
Every id, parent_id, and relation endpoint must match gen-YYYYMMDD-NNN. Duplicates within a single push are rejected.
Each relations[i].from and to must resolve to a generation in the same payload. Cross-snapshot references are rejected.
Only glhub-push/v1 is accepted today. Future versions ship with a documented migration plan before the writer flips.
Three steps from local capture to shared lineage.
Download the latest binary from GitHub Releases. Or build from source with cargo if you prefer.
# macOS (Apple Silicon)
curl -fsSL -o glctl https://github.com/baryonlabs/glctl/releases/latest/download/glctl-aarch64-apple-darwin
chmod +x glctl && sudo mv glctl /usr/local/bin/
# macOS x86_64 / Linux: see Releases for other targets
# Build from source: cargo install --git https://github.com/baryonlabs/glctlBrowser-based GitHub OAuth via loopback. Token saved to ~/.glctl/config — no manual paste.
glctl login
# → Opens browser, GitHub authorizes, token returns to localhost
# → Token saved to ~/.glctl/configFrom any company-scoped lineage repo, push the snapshot to the hosted instance or your self-host.
GLCTL_COMPANY_ID=demo \
GLCTL_DATA_DIR=$HOME/.glctl/data \
glctl push --remote https://glhub.baryon.aiFirst-class adapters for GitHub, GitLab, Codeberg, Forgejo, and self-hosted Gitea. Provider is auto-inferred from any repo URL — no configuration. Webhooks land at /webhooks/{provider} with HMAC verification.
Lineage primitives are Apache-2.0. Multi-user, audit-grade, and on-prem features are commercial. Sentry-style separation, not GitLab-style ee/ subdirectory.
10 generations, all fields populated. See how glhub presents AI evolution as a readable document instead of raw payloads.