Documentation
Architecture, security model, self-hosting, API reference, and plugin SDK for the governed AI application platform.
Evaluating Megapodes for your organization?
Book a demo and get a guided pilot — we walk your team through architecture, security, and deployment.
Documentation is launching with our v1.0 release
The full documentation set below ships alongside v1.0. Until then, the Security Model and Enterprise deployment pages are available now, and our team can walk you through anything else. Get notified using the form below.
Quick Start
Deploy with Docker Compose, connect Ollama for local LLM inference, and build your first app through the gated AI build loop. Covers system requirements, including GPU guidance (~8GB VRAM) for the Ollama host.
Learn MoreArchitecture
System design, microkernel plugin architecture and registries, the declarative metadata model, AI orchestration with LangGraph, and how the fixed, tested runtime interprets app metadata.
Learn MoreSecurity Model
Three-layer authorization — role ACL, forced PostgreSQL Row-Level Security (RLS), and field-level masking applied in SQL — plus the AI security boundary, envelope encryption, and append-only audit logging.
Read the Security ModelSelf-Hosting & Enterprise Deployment
Docker deployment, stateless server and worker replicas, Redis-coordinated cache invalidation, air-gapped operation, SSO/OIDC, and signed dev → test → prod environment promotion.
See Enterprise DeploymentAPI Reference
REST API (/api/v1) with RFC 9457 errors, idempotency keys, and ETags. Covers authentication and JWTs, the metadata build-time API, runtime data CRUD, workflows, AI Employees, and the Model Context Protocol (MCP) Streamable-HTTP endpoint.
API DocumentationPlugin SDK
Develop plugins as npm packages with a manifest and lifecycle: field types, block types, workflow nodes, auth providers, data-source drivers, and AI tools — registered through the same kernel registries the built-ins use.
SDK GuideGet notified when docs go live
Leave your work email and we will notify you when the full documentation ships with v1.0.
Technology Stack
Proven, boring technology chosen for reliability and self-hostability.
Node.js 22
LTS runtime
Fastify 5
HTTP framework
React 18
UI library
Ant Design 5
Component system
TypeScript
Type safety
PostgreSQL 18
Database + pgvector
Redis
Cache & queues
MinIO / S3
Object storage
Ollama
Local LLM inference
LangGraph
AI orchestration
Zod
Schema validation
Turborepo
Monorepo build
Non-Negotiable Invariants
These principles are baked into the architecture. No feature, plugin, or AI action can violate them.
AI emits schema-validated artifacts only
Never executable code. The AI produces declarative metadata validated by Zod schemas — the single source of truth for both generation and runtime.
Human approval for app changes
Preview → approve → apply. No "confirm all." Destructive items need individual confirmation. Every change is audit-logged.
Row-Level Security always on
PostgreSQL Row-Level Security (RLS) is forced on every tenant table. The app role cannot bypass it. Database session context is set per transaction and cleared at the end.
AI obeys permissions
The build-time AI and runtime AI Employees operate under the same three-layer authorization as human users. The rules cannot be bypassed, and no escalation path exists.
Fail-closed
If the model or any gate can't be satisfied, the operation fails loudly. No silent degradation, no fallback to unsafe behavior.
Microkernel — nothing patches core
All extensions happen through kernel extension points in the microkernel plugin architecture. Plugins are add-on packages. Built-in features use the same plugin system, so the design is proven.
See the platform before the docs ship
Our team can walk you through architecture, security, and deployment in a guided pilot.
Self-hosted — your data never leaves your network · Every change human-approved and audit-logged · SSO/OIDC available · We support your security review.
The governed AI application platform. A local AI builds business systems as validated configuration, every change is human-approved, and it all runs on your infrastructure.