Skip to content
LACE
  • v0.1 Current
  • Python
  • TypeScript Soon
GitHub Book a demo

Get Started

LACE Platform and App SDK Documentation

LACE is the AI platform for the governed enterprise — search, knowledge graph, agents, and an app builder on one accountability substrate. This hub covers the public developer surface.

These docs cover the full public developer surface — the lace-app-sdk, the /v1 REST API, and the four product walkthroughs — not just the SDK. In-product docs stay versioned with the release you run; what you read here is the stable, publishable surface.

Start here

If you are building an application, go straight to installing the SDK, the quickstart, and creating your first app (UI / API / CLI). To run it locally, see clone & local dev. If you are evaluating the platform, how LACE is organized, architecture, and deployment options are the shortest path.

The platform

Enterprise Search Cited, permissions-aware answers across every source. Knowledge Graph Bitemporal facts with receipts and conflict detection. Agent Studio Governed agents on chat, email, SMS, and voice. App Builder Prompt → sandboxed build → live, versioned application.

Everything in these docs

Get Started

  • How LACE is organized — How the LACE platform is structured: pipelines, agents, apps, knowledge, and the assistant sharing one tenant-scoped control plane with enforced permissions and evidence.
  • Architecture overview — The LACE control plane, app framework, execution runtimes, knowledge stack, and platform data stores — and how a request flows through them.
  • Deployment options — Deploy LACE to managed cloud, your own VPC, on-premises hardware, or a fully air-gapped enclave with no external AI calls. The governance substrate is identical in every footprint.

lace-app-sdk

  • Install — Install lace-app-sdk, the public contract for building governed enterprise AI applications on LACE. Requires Python 3.12 or later.
  • Quickstart — Build a governed LACE application from four small files: the manifest, a data collection, a tool provider, and an agent definition.
  • Create a new app — Create a LACE app three ways: from the App Builder UI, from the REST API, or from the lace-app CLI. Same manifest, same contracts, three entry points.
  • Clone & local dev — Clone a graduated app repo with lace-app clone, run it locally with lace-app dev, and understand the compose stack, emulator, and .lace state.
  • What gets installed — The file tree, manifests, generated scaffolding, templates, and dependencies that a new LACE app installs — and what each file owns.
  • The SDK itself — How the SDK re-exports live lace.* code, the base and runtime installs, versioning, and the full module map.
  • Manifest — LaceAppManifest declares an app's routes, tools, pipelines, agents, data collections, UI, schedules, and migrations. LACE discovers manifests and wires providers at bootstrap.
  • Data collections — Declare typed, tenant-scoped data collections with JSON Schema and serve them through the governed AppDataService, with automatic admin panels and audited blob storage.
  • Tools — Expose custom tools to LACE agents with AppToolProvider. Tools run in constrained runtimes with trust tier and permissions checked before every dispatch.
  • Agents & skills — Declare an AgentDefinition to run a real agentic loop in your sidecar, with platform-enforced capabilities, loop guards, reusable skills, and human approval gates.
  • Pipelines — Declare pipelines as versioned, typed graphs of steps with checkpoints, retries, idempotency, and dead-letter handling — compiled the same way the platform executes them.
  • Routes & UI — Expose HTTP routes and federated runtime-UI modules from your app sidecar. Routes mount at /apps/<app_id>/api/*, UI at /ui/manifest.json.
  • Testing & proof lanes — Validate apps locally with lace-app test: typecheck, build, route smoke, tool invocation, permission, and AppData contract lanes.
  • Publishing & releases — Seal, push, deploy, and version your app. Releases are immutable, content-addressed, and one rollback away.
  • CLI reference — The lace-app CLI: auth, create, clone, dev, test, push, deploy, open, logs, status, releases, and secrets — with flags, env vars, and examples.
  • Module reference — Reference for every lace-app-sdk module: manifest, data, tools, agents, skills, pipelines, routes, connectors, memory, policy, migrations, ui, and testing.

Platform Guide

  • Datasets & connectors — Connect SharePoint, Google Drive, OneDrive, Dropbox, Notion, Slack, email, S3, databases, and websites, with scheduled sync, change reconciliation, and permission snapshots.
  • Enterprise Search — Product walkthrough for Enterprise Search: three-leg retrieval, RRF fusion, ACL enforcement, GPU reranking, and cited answers that never leak what you cannot see.
  • Knowledge Graph — Versioned executable ontology contracts, candidate-first evidence-gated extraction, reversible identity resolution, and bitemporal assertions that supersede rather than erase.
  • Agent Studio — Product walkthrough for Agent Studio: design the tool surface, write the system prompt, wire channels, enforce policy, and ship to web, email, SMS, and voice.
  • Workflows & Pipelines — Versioned graphs of typed steps with checkpoints, retries, idempotency, and dead-letter handling. Workflow Studio compiles the same contract the API executes.
  • App Builder — Describe an application and the builder plans it, constructs it in a sandbox, validates it through proof lanes, and publishes it as an isolated sidecar with versioned releases.
  • Governance & security — Permission scopes, resource ACLs, budgets, usage limits, approval gates, structured logs, OpenTelemetry traces, and per-call model attribution for every AI operation.

API & Platform

  • REST API & other SDKs — The /v1 REST API, lace-app-sdk (Python), frontend SDKs, CLI, webhooks, events, and MCP — auth, versioning, pagination, and errors.
  • Authentication — API keys, JWTs, tenant scoping, RBAC scopes, resource ACLs, and the policy checks that guard every route, tool, and retrieval path.
  • Observability — Structured logs, OpenTelemetry traces, per-call model attribution, pipeline replay, agent session inspection, and cost dashboards.