Governance foundation
- AI assets and versioning for prompts, models, datasets, and test providers
- Multi-tenancy, access control, settings, and durable records
- Policy rules, explainable decisions, and governance ontology / lineage
Roadmap
AIGP is in an adoption-led phase. We are prioritising usability, integrations, operational reliability, and real-world feedback over adding features for their own sake.
Vision
AIGP helps teams govern AI systems across their lifecycle without taking over the systems that run them. It observes execution, records important changes, evaluates behaviour, applies rules to decisions, safely replays evidence, and preserves the history that explains what happened.
The aim is a durable governance layer that remains useful as providers, runtimes, and delivery tools change. AIGP keeps the evidence and decisions around AI work clear and reviewable.
What AIGP Owns
- Governance assets and versions
- Evaluation evidence, policies, and decisions
- Runtime execution evidence and findings
- Controlled Replay and Causal Audit records
- Durable governance work, lineage, and audit history
- Stable integration contracts
What AIGP Does Not Own
- Model and agent execution
- Production deployment
- Infrastructure orchestration
- CI/CD execution
- Application runtime orchestration
AIGP governs the evidence and decisions around AI systems; your runtime executes them and your delivery systems deploy them.
Shipped
These capabilities are available today. They are grouped by what they let you do, rather than by the order in which they were built.
Evaluation and change
- Evaluation framework, experiments, comparison history, and drift analysis
- Saved workflow replay with linked evidence and review history
- Governed artifacts and approval-aware change records
Agent governance
- Agents Runtime for safe execution evidence and timelines
- Runtime Findings for repeat conditions that need review
- Intervention Policies, Controlled Replay, and Causal Audit
Platform and integrations
- Durable Job Execution Control Plane
- REST API, MCP, audit, persistence, and observability
- Public Runtime Extensibility through a stable, independently versioned Plugin API
Current focus
We have reached the point where proving the platform is useful, understandable, and dependable matters more than expanding its feature surface.
- Simplify installation and the first-run experience
- Improve guided onboarding and product documentation
- Make Agents Runtime, Replay, and Causal Audit easier to understand and operate
- Validate external runtime integrations through the public Plugin API
- Improve examples, reference implementations, and reproducible demos
- Gather OSS usage feedback and privacy-conscious telemetry
- Harden the capabilities people actually use
- Reduce operational friction before expanding the platform further
Under consideration
Future work will be driven by adoption, integration demand, and operational evidence—not a fixed feature commitment. These are directions we are evaluating, not promises or dated delivery plans.
- Additional runtime integrations
- Developer SDKs
- A connector ecosystem
- Governance and compliance packs
- Managed deployment options