> ## Documentation Index
> Fetch the complete documentation index at: https://lightsage.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Onboard your AI agent to Lightsage

> Give your coding agent Lightsage skills, connected tools, workflows, and current documentation.

Give your coding agent reusable Lightsage instructions first, then connect it to your workspace and provide workflows and current documentation.

## Agent Skills

Install three Lightsage skills for prompt-run analysis, eval diagnosis, and CLI operation.

> **[Install Agent Skills](/docs/agent-onboarding/agent-skills)**
>
> Add all three skills or choose the ones that fit your workflow.

## Lightsage MCP

Use MCP when your agent runs in a hosted connector such as ChatGPT or Claude, or cannot run a shell. Use OAuth sign-in where supported, or configure the `X-Lightsage-Api-Key` header.

> **[Connect through Lightsage MCP](/docs/agent-onboarding/lightsage-mcp)**
>
> Choose your client, connect to the hosted MCP server, and verify access.

## Lightsage CLI

Use the CLI for coding agents with shell access and for workflows that poll, filter, or script structured output. Create an API key under **Settings > API Keys** and store it as `LIGHTSAGE_API_KEY`.

> [!WARNING]
>
> Do not put the key in shared project config, screenshots, issue reports, or public repositories.

> **[Set up the Lightsage CLI](/docs/agent-onboarding/lightsage-cli)**
>
> Install the CLI, authenticate, and verify access.

## Workflows

> **[Use agent workflows](/docs/agent-onboarding/workflows)**
>
> Turn prompt-run and eval evidence into prioritized actions.

## Docs for agents

Give your agent current Lightsage documentation instead of relying on its training data.

> **[Give your agent the docs](/docs/agent-onboarding/docs-for-agents)**
>
> Use page-level Markdown or the complete documentation index.

## Verify access

Give your coding agent this prompt:

```text theme={null}
Check that you can access my Lightsage workspace. Use the connected Lightsage
MCP server if available; otherwise use the Lightsage CLI. List my available
evals and agents, and report any missing prerequisites.
Do not create resources or start eval runs.
```

This checks CLI / MCP authenticated access, not whether the skill is installed or loaded. An empty eval list is a valid result for a new organization. To check skill installation, use your coding agent’s installed-skills list where available.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.