Overview
Documentation for `@weave-kit/engine` — let AI agents operate your data, safely.
Let AI agents operate your data — safely.
The pitch and quick start live in the repository README.
Start here
- Getting started — scaffold, migrate, run, call the API
- Connect an agent — a local first run in five minutes
- Scaffolding a preset?
--type=agent|governance|service→ Project types
By goal
- Put an agent on an existing database → Existing CRM → MCP
- Model your data → Schema · Named enums · Formulas
- Lock it down → RBAC · Identity · Multi-tenancy · Audit
- Govern a lifecycle → Workflow — role gates, approvals, timeouts · Execution & evidence
- Run it in production → Deployment · Observability · Release checklist
Guides
- Project types — the three presets and what each scaffolds
- Model — schema, formulas, git-versioned metadata
- Access — RBAC, identity and multi-tenancy
- Agents — MCP, custom tools & guardrails, approvals, execution & evidence
- Workflow — per-object approval chains
- Automation — script hooks, quotas and the query budget
- Integration — live events, inbound events, proxy, tunnel, GraphQL
- Platform — CLI, audit, i18n, observability
Operations
- Deployment — containerize, reverse-proxy, TLS
- Release checklist — the maintainer checklist for cutting an engine release
Practices
- Agent — wiring an existing database into an agent surface
- Governance — customer user stores, governed lifecycles
- Service — service-shaped backends
- Business — object/formula backends
- Modeling — shared vocabularies and reusable definitions
Reference
- API — public API tiers and the dependency budget
- Schema — custom field types
- Contract freeze — what is frozen at 1.0 and the deprecation policy
- Enterprise seams — the stable extension points behind the enterprise layer
- Migrating to 2.0 — the breaking changes (contract package,
scope, retired groundwork)