Crystal Lattice Docs

Crystal Lattice is a plugin-driven AI agent platform for running local-first, provider-agnostic agent workflows across your computer, phone, and custom automation.

The platform includes a Python SDK, a terminal and HTTP runtime, mobile and desktop clients, provider integrations, and a first-party plugin system for tools, features, provider behavior, application actions, and model-family harnesses.

Start Here

Package documentation: choose the SDK, application, component, or service that owns the part of Crystal Lattice you want to use.
Application configuration: configure agents, providers, plugins, tools, placeholders, and runtime policy.
Current plugin docs: build provider, feature, application, and tool plugins using the current plugin documentation set.
Built-in plugins: browse generated reference pages for bundled plugin packages.
Default configuration: inspect the generated default application configuration.
Event stream types: understand application events exposed by HTTP polling and bridge streaming.
Development documentation: reusable standards, Crystal Lattice repository policy, and cross-package target specifications.

Developer Reference

Reusable documentation standard: package-neutral documentation ownership, document roles, and discoverability rules.
Reusable specification standard: target specifications, live requirements, work tracking, and evidence.
Crystal Lattice repository policy: local package profiles, plugin documentation, task workflow, and publication rules.
AgentCore API: core session, message, request, streaming, and tool orchestration APIs.
Core SDK guide: hand-written overview of the functional SDK surface.
Types and protocols: immutable core types and plugin protocols.
Provider wrapper: provider adapter API reference.
OpenAI-compatible provider: reference provider implementation details.
Tool host protocol: protocol reference for external tool hosts.

Project Links

Use the Website Home menu item to return to the Crystal Lattice website.
GitHub repository