Change Governance Protocol
The first standardized, vendor-neutral protocol for governing software change. Actor-agnostic, transport-independent, audit-first.
The governance layer for software change
As AI agents and CI systems generate more code, deciding what should ship becomes the hardest problem. Relicta governs change — before it reaches production.
The Problem
Code is cheap. Changes are frequent. Risk is hidden. Trust erodes.
Human-written code. Human reviews. Human pace.
Agent-generated change. CI automation. Overwhelmed humans.
Governance. How do we trust the machine to ship?
The Solution
Relicta sits between humans, agents, and CI systems. It decides what changes, how risky it is, who approves, and how it's communicated.
How It Works
One command. Five stages. Full audit trail.
Analyze commits, detect breaking changes, calculate risk score, suggest version.
Apply semantic version. Create git tag.
Generate AI-powered release notes. Engineering, product, executive, and customer variants.
Policy check. Auto-approve low-risk, require humans for high-risk. Audit hash chain entry.
Tag, changelog, plugins, Sigstore attestation. Done.
$ relicta release
plan 47 commits since v1.3.0 · 2 breaking, 8 features, 12 fixes
Risk: 0.58 (medium) · Blast radius: 3/7 packages
Suggesting: v2.0.0 (major)
bump Version set to v2.0.0
notes Release notes generated via Claude
4 audience variants: engineering, product, executive, external
approve Policy: risk > 0.5 requires approval
Auto-approved: actor trust score 0.94
Audit: SHA-256 hash chain entry recorded
publish ✓ GitHub release created
✓ npm package published
✓ Slack #releases notified
✓ in-toto attestation signed via Sigstore
Done. v2.0.0 released with full governance trail. Differentiators
Other tools automate releases. Relicta governs them.
The first standardized, vendor-neutral protocol for governing software change. Actor-agnostic, transport-independent, audit-first.
The only release tool that's an MCP server. AI agents can plan, assess risk, approve, and publish releases natively.
7-factor weighted risk calculation that learns from past releases and incidents. Gets smarter every time you ship.
Maps changed files to impacted packages, builds dependency graphs, and quantifies change scope before you release.
One release, four stories. Engineering gets the diff. Product gets highlights. Executives get impact. Customers get the upgrade guide.
Composable governance rules with full logic support. Define approval gates, team routing, and auto-approval criteria as code.
Tracks actor reliability, incident correlation, and risk patterns over time. Your release intelligence compounds.
in-toto v1 attestation signed via Sigstore. Proves not just what was built, but why it was approved and by whom.
Comparison
| Capability | semantic-release | release-please | goreleaser | LaunchDarkly | Relicta |
|---|---|---|---|---|---|
| Semver automation | ✓ | ✓ | ✓ | — | ✓ |
| AI release notes | — | — | — | — | 5 providers |
| Risk scoring | — | — | — | Basic | 7-factor + learning |
| Approval workflows | — | — | — | — | Policy DSL |
| Audit trail | — | — | — | Flags | Cryptographic |
| MCP server | — | — | — | — | Native |
| Blast radius | — | — | — | — | ✓ |
| Supply chain attestation | — | — | Partial | — | in-toto + Sigstore |
| Single binary | — | — | ✓ | — | ✓ |
| Open source | MIT | Apache | MIT | — | MIT |
Trust & Security
Runs locally or in your CI. No SaaS. No cloud dependency. Your code, your keys, your infrastructure.
Every governance decision is recorded in an immutable SHA-256 hash chain. Tamper-evident by design.
in-toto v1 statements signed via Sigstore prove what was released, why it was approved, and by whom.
Plugins run in isolated processes with capability-based restrictions on filesystem, network, and environment access.
Full software bill of materials for supply chain transparency and compliance.
API keys and tokens are automatically redacted from all output and logs. Enabled by default.
Production Ready
Start with release automation. Grow into full change governance.
The Evolution
Built For
Standardize release workflows across hundreds of services. One config, consistent governance, full observability. Monorepo support with independent, lockstep, and hybrid versioning.
Regain visibility without slowing anyone down. Risk scores surface what needs attention; auto-approval handles what doesn't. Real-time dashboard with WebSocket streaming.
Immutable hash-chain audit log. SLSA in-toto attestation with Sigstore signing. SBOM generation. OIDC/SSO with role-based access.
CGP protocol governs agent-initiated changes. Policy DSL defines what agents can auto-approve. Actor trust scores track reliability over time.
AI-Native
MCP is the industry standard for connecting AI agents to tools. Relicta is a native MCP server — not a wrapper, not a plugin.
relicta.plan Analyze commits & suggest versionrelicta.bump Apply semantic versionrelicta.notes Generate release notesrelicta.evaluate CGP risk evaluationrelicta.approve Governance gaterelicta.publish Execute releaserelicta.blast_radius Impact analysisrelicta.validate_release Pre-flight checksrelicta://state Release state machinerelicta://config Project configurationrelicta://commits Pending commitsrelicta://changelog Generated changelogrelicta://risk-report Risk assessmentLive release status at a glance
Step-by-step release progress
Visual risk scoring breakdown
Classified commit analysis
Interactive governance gates
Impact analysis across packages
Open Source
We believe the governance layer must be open. No vendor lock-in. Full transparency.
One binary. Zero cloud dependencies. Full governance trail.
brew install relicta-tech/tap/relicta