Integrations
Integration categories, not a vendor logo wall
This catalog describes the generic categories of external system the kernel is designed to connect through — source control, CI/CD, cloud infrastructure, model providers, and more — with their authentication model, direction, and constraints. No specific provider, partner, or vendor is named here.
No integration category below is available for use through this website today. Source control, CI/CD, observability, and identity are planned; the remaining categories are vision-stage. A named partner integration will only be added once a real, governance-approved partnership exists.
Integration category catalog
Every entry uses the same generic categories declared in the integration schema: source control, CI/CD, cloud provider, model provider, agent framework, editor/IDE, observability, identity, storage, database, and messaging.
Agent framework
Frameworks for building and orchestrating AI agents and tool-using systems
- Direction
- read, write
- Authentication models
- api-key, oauth
- Agent actions must pass control-plane gates
CI/CD
Continuous integration and continuous deployment pipelines
- Direction
- read, write
- Authentication models
- api-key, oauth, token
- Pipeline secrets never exposed to agents
Cloud provider
Infrastructure and compute platforms for deployment and scaling
- Direction
- read, write
- Authentication models
- service-account, oauth, iam-role
- Cloud resources gated by approval workflow
Database
Structured data storage and query systems
- Direction
- read, write
- Authentication models
- connection-string, iam-role
- Read-only access for agents unless explicitly approved
Editor / IDE
Development environments and code editors
- Direction
- read, write
- Authentication models
- local, oauth
- Editor integration requires local agent process
Identity
Authentication and identity management systems
- Direction
- read
- Authentication models
- oauth, saml, api-key
- Identity operations require human approval
Messaging
Communication and notification platforms
- Direction
- read, write
- Authentication models
- api-key, oauth, webhook
- Outbound messaging gated by evidence
Model provider
Large language model and AI inference services
- Direction
- read
- Authentication models
- api-key
- Provider-neutral by design; no vendor lock-in
Observability
Monitoring, logging, and tracing systems
- Direction
- read, write
- Authentication models
- api-key, oauth
- Observability data read-only for agents
Source control
Version control and code repository systems
- Direction
- read, write
- Authentication models
- api-key, oauth, ssh
- All changes require evidence-bound commits
Storage
Object and file storage systems
- Direction
- read, write
- Authentication models
- api-key, iam-role
- Storage access bounded by provenance gates
Fail-closed integration boundary
An integration category being listed here does not grant standing authority. Every integration is scoped to explicit, time-limited credentials and mission-level authorization; if authorization cannot be verified, the integration fails closed rather than defaulting to access.
- No integration self-authorizes a production or irreversible effect.
- No partner relationship or supported integration is implied by category inclusion alone.
- Read access to a category does not imply write access; each is scoped independently.