# AgentLattice

> TypeScript framework for building coordinated agent systems with tools, skills, tracing, and teams.

AgentLattice runs an agent loop against any Anthropic-compatible model, executes
tools you define, and coordinates several agents through supervisor delegation or
durable mailbox teams. Install it with `npm install agent-lattice zod`.

This documentation describes agent-lattice v0.23.0.

If the package is already installed, check the installed version before writing
code against these pages, and upgrade if it is older. Unknown options are ignored
rather than rejected, so calling a newer API on an older install succeeds with the
feature silently doing nothing — there is no error to catch. Install or upgrade
with `npm install agent-lattice@0.23.0`.

Every link below serves clean Markdown. Fetch the page that matches the task at
hand, or read llms-full.txt for the whole documentation in one request.

## Start

- [Quickstart](https://docs.claude-code-sdk.com/quickstart.md): Install AgentLattice and run a minimal agent.
- [Configuration](https://docs.claude-code-sdk.com/configuration.md): Configure models, providers, turn limits, and request controls.

## Core Concepts

- [Agent Loop](https://docs.claude-code-sdk.com/concepts/agent-loop.md): How the SDK turns model responses and tool calls into an agent session.
- [Streaming Events](https://docs.claude-code-sdk.com/concepts/streaming-events.md): Consume stable SDK messages and raw model stream events.
- [Context Tracing](https://docs.claude-code-sdk.com/concepts/context-tracing.md): Record context traces for agent runs, tool calls, and delegated team work.
- [Tools](https://docs.claude-code-sdk.com/concepts/tools.md): Define custom tools with schemas, parsed input, and typed handlers.
- [Hooks](https://docs.claude-code-sdk.com/concepts/hooks.md): Rewrite tool results and outgoing model requests as they cross the agent loop.
- [MCP](https://docs.claude-code-sdk.com/concepts/mcp.md): Connect MCP server tools to the agent loop.
- [Skills](https://docs.claude-code-sdk.com/concepts/skills.md): Reuse instruction bundles without depending on the Claude Code runtime.
- [Delegation and Teams](https://docs.claude-code-sdk.com/concepts/multi-agent.md): Choose between one-shot supervisor delegation and persistent mailbox teams.
- [Supervisor Delegation](https://docs.claude-code-sdk.com/concepts/supervisor-delegation.md): Use one-shot AgentLike child agents as tools.
- [Mailbox Team](https://docs.claude-code-sdk.com/concepts/mailbox-team.md): Build persistent teams with lead agents, member inboxes, replies, and resumable work.
- [Built-in Tools](https://docs.claude-code-sdk.com/concepts/built-in-tools.md): Agent workspace tools and manual workspace tool construction.
- [Permissions](https://docs.claude-code-sdk.com/concepts/permissions.md): Intercept tool calls before execution and apply host policy.

## Reference

- [Public API](https://docs.claude-code-sdk.com/reference/public-api.md): Stable exports exposed by the SDK package.
- [Provider Compatibility](https://docs.claude-code-sdk.com/reference/provider-compatibility.md): How SDK thinking, reasoning-effort, and structured-output options behave on Anthropic-compatible providers such as DeepSeek and Kimi.

## Optional

- [Full documentation in one file](https://docs.claude-code-sdk.com/llms-full.txt): every page above concatenated, for agents that prefer one large read over several small ones.