llms.txt@site/static · git:20260918.31a6393 · 2026-09-18 · sha256 f22838f78d3409d9
llms.txt@site/static git:20260918.31a6393A
Immutable. This exact content never changes and is served at /api/v1/blob/f22838f78d3409d9.
# Agent Layer > Agent Layer is a repo-local control plane for coding agents. It provides asynchronous cross-provider Agent Dispatch, Git-backed portable Agent Skills, shared MCP and approval configuration, and DeltaSelect benchmark selection. The canonical source repository is https://github.com/conn-castle/agent-layer. Documentation pages are server-rendered for browsers and also published as Markdown at the `.md` URLs below for agents and text-only clients. ## Start here - [Agent Dispatch](https://agent-layer.dev/docs/agent-dispatch.md): Delegate asynchronous work to Claude, Codex, Grok, or Antigravity through seven MCP tools or matching CLI commands. - [Import Agent Skills from Git](https://agent-layer.dev/docs/skill-imports.md): Import, edit, update, reconcile, and contribute portable Agent Skills without losing project changes. - [Getting started](https://agent-layer.dev/docs/getting-started.md): Install, initialize a repository, sync configuration, and launch a coding agent. - [Reference](https://agent-layer.dev/docs/reference.md): Authoritative configuration, CLI, Agent Dispatch, benchmark runner, and skill-import behavior. - [DeltaSelect](https://agent-layer.dev/deltaselect): Select informative AI-agent benchmark tasks within a fixed evaluation budget. ## Core concepts - [Overview](https://agent-layer.dev/docs/overview.md): Product overview and documentation map. - [Concepts](https://agent-layer.dev/docs/concepts.md): Source-of-truth model, safety boundaries, MCP servers, dispatch, and version pinning. - [Skills](https://agent-layer.dev/docs/skills.md): Agent Skills standard, built-in workflows, authoring, and projection. - [Skills approach](https://agent-layer.dev/docs/skills-approach.md): Agent Layer's skill architecture and workflow boundaries. - [Security](https://agent-layer.dev/security): Network access, secret handling, generated files, and trust boundaries. - [Install](https://agent-layer.dev/install): Supported installation methods for macOS and Linux. ## Additional resources - [FAQ](https://agent-layer.dev/faq): Direct answers about Agent Dispatch, skill imports, clients, MCP, security, and upgrades. - [Troubleshooting](https://agent-layer.dev/docs/troubleshooting.md): Common errors and actionable fixes. - [Changelog](https://agent-layer.dev/changelog): Release history. - [GitHub repository](https://github.com/conn-castle/agent-layer): Canonical code, README, issues, and source documentation.