Welcome
Adapstory is a GenAI-native, plugin-first LMS platform. Build plugins that extend the runtime, data model, and AI gateway.
These docs are for developers building and deploying plugins on the Adapstory platform.
Platform identity
Adapstory is GenAI-native (LLM is the core, not an add-on) and plugin-first (everything that can be a plugin must be a plugin). Core is minimal: BC-10, BC-02, BC-15, BC-16, BC-19, BC-01. All business value ships as plugins.
Start here
Platform overview
How the bounded contexts fit together, what the plugin runtime does, and where your code plugs in.
Plugin quickstart
From zero to a running plugin in ~30 minutes. Manifest, handler, local run, publish.
Plugin manifest reference
Every field in plugin.yaml — identity, MCP tools, model configs, guardrails, permissions.
Lifecycle & deployment
Install, update, rollback, and tenant isolation. How Kyverno verifies your signed image.
What you'll build
Plugins on Adapstory can:
- Extend the Data Model Engine (BC-15) with custom entities, validations, and event reactions.
- Register MCP tools with the Plugin Gateway (BC-01) so LLMs can invoke your capabilities.
- Run isolated tenant logic under the Multi-Tenant Runtime (BC-19) with per-tenant data walls.
- Ship AI features with versioned guardrails, token-budget enforcement, and A/B-tested model configs.
Contract guarantees
- Plugins are signed with Cosign and accompanied by SBOM + SLSA provenance. Kyverno enforces this at admission.
- Plugin-to-core communication is Kafka + CloudEvents or REST. No shared databases.
- LLM access is only via the LLM Gateway — direct provider calls are forbidden.
- Secrets flow via Vault → External Secrets Operator → Kubernetes Secret. Never in manifests.
Help & contributions
- Source: github.com/adapstory/Adapstory-GitOps (
infra/docs/content/) - Issues: github.com/adapstory/Adapstory-GitOps/issues
- Edit a page: every doc has an "Edit on GitHub" link in the top-right.