OpenAegis is an open-source control layer for enterprise AI agents in regulated environments.
In plain English: the model can suggest work, but OpenAegis decides what is allowed, what must be approved, and what must be recorded as evidence.
OpenAegis is built for organizations where data leakage is a business and compliance failure, not just a bug.
| Option | Good at | Gap in regulated enterprise use | OpenAegis advantage |
|---|---|---|---|
| Generic agent frameworks | Fast prototyping | Usually weak policy, approvals, and audit replay | Policy enforced outside model + evidence chain |
| Vendor-specific copilots | Convenience | Vendor lock-in and limited control portability | Vendor-neutral model broker |
| DIY scripts | Custom logic | Hard to prove safety and control coverage | Built-in approvals, audit, incidents, simulation |
| OpenAegis | Controlled automation | Early-stage project, expanding coverage | Security-first defaults and executable proof |
- Policy and approvals are enforced outside the model.
- High-risk live actions can be blocked pending human approval.
- Security and compliance teams get evidence IDs for replay and review.
- Sensitive routing can require zero-retention providers.
- Multi-tenant context is explicit and carried across requests.
- Simulation mode exists before live execution.
- Plugin/extension arm supports OAuth, API key, service-principal, and key-pair onboarding patterns.
OpenAegis now includes a broader plugin lifecycle through tool-registry:
- Catalog:
GET /v1/tools - Plugin instances:
GET/POST /v1/plugins/instances - OAuth authorization:
POST /v1/plugins/instances/{id}/authorize - Connection test:
POST /v1/plugins/instances/{id}/test
Included connector families now cover AWS, Databricks, Fabric, Jira, Confluence, OpenAI, Anthropic, Google, Azure OpenAI, Airbyte, Airflow, Trino, Superset, Metabase, Grafana, Kafka, NiFi, Dagster, n8n, plus healthcare and operations connectors.
Run the commercial showcase harness:
npm run showcase:projectsThis validates five commercially relevant projects with live API execution:
- SecOps Runtime Guard
- Revenue Cycle Copilot
- Supply Chain Resilience
- Clinical Quality Signal
- Board Risk Cockpit
Artifact:
docs/assets/demo/commercial-projects-showcase-report.json
In the Admin Console, open http://127.0.0.1:4273/projects to see:
- seeded operational tables per pack
- live-evaluated policy scenario outcomes
- one-click secure baseline policy presets
- step-by-step walkthrough cards with evidence expectations
Guide: docs/projects/STEP-BY-STEP-DEMO.md
OpenAegis now includes a guided Policy Studio in the Security Console.
- Edit beginner-safe controls with plain-language explanations.
- Run Preview Impact to see
ALLOW / REQUIRE_APPROVAL / DENYchanges. - Run Explain Impact to see risk score delta and per-control safety guidance.
- Review warnings before saving.
- Use LLM Copilot to review and suggest safer settings.
- Apply the policy profile (break-glass fields are required for blocking-risk changes).
See detailed guide: docs/policy-studio.md
OpenAegis supports a hardened gateway auth path using auth-service introspection.
Set:
OPENAEGIS_AUTH_INTROSPECTION_URLOPENAEGIS_REQUIRE_INTROSPECTION=true
Secure default:
- Insecure demo login/token mode is disabled unless explicitly enabled with
OPENAEGIS_ENABLE_INSECURE_DEMO_AUTH=true.
In this mode, bearer tokens are introspected and write operations enforce tenant scope from token claims.
Run the executable trust proof harness:
npm run proof:trust-layerThis runs three distinct examples:
- Healthcare discharge orchestration (policy + approval + audit replay)
- Finance operations guardrails (runtime blocks + idempotent retries)
- SecOps containment (break-glass + kill-switch + immutable chain)
Artifact:
docs/assets/demo/trust-layer-proof-report.json
The included pilot demonstrates a Hospital Discharge Readiness Assistant:
- Read patient context from FHIR + SQL connectors.
- Route model inference based on sensitivity policy.
- Block high-risk outbound follow-up actions pending human approval.
- Capture immutable audit/evidence for every major action.
| Claim | Where it is proven |
|---|---|
| Policy gates enforced outside model | backend/services/api-gateway/src/index.test.ts |
| High-risk live approval gating | backend/services/api-gateway/src/index.test.ts |
| Break-glass required for blocking policy downgrades | backend/services/api-gateway/src/index.test.ts |
| Policy impact explainability and advisor fix path | /v1/policies/profile/explain + Security Console |
| Copilot guidance for policy edits | /v1/policies/profile/copilot + Security Console |
| End-to-end pilot still operational | npm run smoke:pilot |
| Commercial evidence output | docs/assets/demo/commercial-proof-report.json |
All screenshots are generated from live route interactions:
- Setup Center:
docs/assets/screenshots/commercial-setup.png - KPI Dashboard:
docs/assets/screenshots/commercial-dashboard.png - Project Packs:
docs/assets/screenshots/commercial-projects.png - Sandbox Proof Review:
docs/assets/screenshots/commercial-sandbox-proof.png - Pack Guide (SecOps Runtime Guard):
docs/assets/screenshots/commercial-project-guide-secops-runtime-guard.png - Pack Guide (Revenue Cycle Copilot):
docs/assets/screenshots/commercial-project-guide-revenue-cycle-copilot.png - Pack Guide (Supply Chain Resilience):
docs/assets/screenshots/commercial-project-guide-supply-chain-resilience.png - Pack Guide (Clinical Quality Signal):
docs/assets/screenshots/commercial-project-guide-clinical-quality-signal.png - Pack Guide (Board Risk Cockpit):
docs/assets/screenshots/commercial-project-guide-board-risk-cockpit.png - Commercial Readiness:
docs/assets/screenshots/commercial-readiness.png - Integration Hub:
docs/assets/screenshots/commercial-integrations.png - Identity & Access:
docs/assets/screenshots/commercial-identity.png - Security Console (Policy Studio):
docs/assets/screenshots/commercial-security.png - Approval Inbox:
docs/assets/screenshots/commercial-approvals.png - Incident Review Explorer:
docs/assets/screenshots/commercial-incidents.png - Audit Explorer:
docs/assets/screenshots/commercial-audit.png - Workflow Designer:
docs/assets/screenshots/commercial-workflow.png - Simulation Lab:
docs/assets/screenshots/commercial-simulation.png - Admin Console:
docs/assets/screenshots/commercial-admin.png
npm install
npm run typecheck
npm run test
npm run build
npm run smoke:pilot
npm run validate:test-surface
npm run security:regression
npm run proof:trust-layer
npm run sandbox:proof
npm run audit:codebase
npm run trust:pack
npm run trust:audit
npm run run:tool-registryRun pilot demo output:
node tools/scripts/pilot-demo.mjsConsole UX starts at:
http://127.0.0.1:4273/setupfor guided onboardinghttp://127.0.0.1:4273/projectsfor five commercial project packshttp://127.0.0.1:4273/integrationsfor Databricks/Fabric/Snowflake/AWS setuphttp://127.0.0.1:4273/identityfor user and role administration
Capture screenshots:
npm run screenshots:commercialGenerate and verify all five live sandbox packs (Trino/Airflow/FHIR/OpenSearch/OPA/MinIO, etc.):
npm run sandbox:proofOne-command local pilot (readiness + KPI + security pack):
npm run pilot:localLaunch the 14-day trust proof challenge:
npm run challenge:launchUse these checks before a release candidate is promoted:
npm run typecheck
npm run build
npm run test
npm run validate:test-surface
npm run validate:infra
npm run security:regression
npm run smoke:pilot
npm run proof:commercial
npm run proof:trust-layer
npm run audit:codebase
npm run trust:pack
npm run trust:audit
npm run audit:commercial
npm run load:commercial
npm run chaos:commercial
npm run readiness:gateExpected pass results:
npm run typecheckexits 0npm run buildexits 0npm run testexits 0npm run validate:test-surfaceexits 0 and verifies every workspace has executable testsnpm run validate:infraexits 0 and validates Docker/Kubernetes/Helm packagingnpm run security:regressionexits 0 and writesdocs/assets/demo/security-regression-report.jsonnpm run smoke:pilotexits 0 and refreshes the pilot evidence bundlenpm run proof:commercialexits 0 and writesdocs/assets/demo/commercial-proof-report.jsonnpm run proof:trust-layerexits 0 and writesdocs/assets/demo/trust-layer-proof-report.jsonnpm run audit:codebaseexits 0 and writesdocs/assets/demo/codebase-line-audit-report.jsonnpm run trust:packexits 0 and updatesdocs/assets/enterprise-trust-pack/latestnpm run trust:auditexits 0 and writesdocs/assets/demo/enterprise-trust-pack-audit-report.jsonnpm run audit:commercialexits 0 and writesdocs/assets/demo/commercial-audit-report.jsonnpm run load:commercialexits 0 and writesdocs/assets/demo/load-test-report.jsonnpm run chaos:commercialexits 0 and writesdocs/assets/demo/chaos-report.jsonnpm run readiness:gateexits 0, writesdocs/assets/demo/readiness-gate-report.json, and enforces >= 98%
Expected proof report fields:
summary.status = PASSsummary.failedClaims = 0summary.scorePercent = 100- readiness gate
summary.status = PASS - readiness gate
summary.scorePercent >= 98
If any one of those checks fails, the release is a no-go.
- Platform blueprint
- MVP plan
- Commercial readiness
- Project packs value brief
- Why OpenAegis
- CISO decision brief
- Enterprise trust pack
- Adoption playbook
- OpenClaw adoption matrix
- Security evidence pack guide
- Licensing model
- Trademark policy
- 14-day trust proof challenge
- Design-partner pilots
- Commercial project packs
- Hospital production gate
- SRE runbook
- Hardening controls matrix
- Policy Studio guide
- Pilot runbook
- Smoke and pilot report
- Operator manual
- Training manual
- FAQ
- Setup support guide
- Top-20 language packs
OpenAegis uses dual licensing:
- AGPL-3.0-only: LICENSE
- Commercial terms: LICENSE-COMMERCIAL.md
Validated in this repository:
npm run typecheckpassesnpm run testpassesnpm run buildpassesnpm run smoke:pilotpasses
OpenAegis is early and evolving. Contributions are welcome for runtime hardening, connectors, policy packs, observability, and localization quality.
