# LLM install guide for ClearSky AI
These are official instructions for AI agents that need to install ClearSky AI for a user. Use the matching client flow below, complete browser auth when needed, and confirm the server is ready before moving on.

## Quick start
fetch https://staging.clearskyai.co.uk/llms.txt to install ClearSky AI
Canonical install doc: https://staging.clearskyai.co.uk/llms.txt

## Agent routine
- Identify whether the user is on Codex, Claude Code, ChatGPT, or Claude.
- Use the matching ClearSky AI install flow below instead of inventing one.
- Complete any browser sign-in or consent handoff that appears.
- Verify the connection inside the client after setup.

## MCP server
https://staging-api.clearskyai.co.uk/mcp

## CLI installs
### Claude Code
Install command:
claude mcp add --transport http clearsky-ai --scope user https://staging-api.clearskyai.co.uk/mcp
Steps:
1. Run the command once in any terminal where `claude` is installed.
2. When Claude Code first calls ClearSky AI, approve the browser auth flow back into ClearSky AI.
3. Use `/mcp` inside Claude Code any time you want to verify the server is connected.
Docs:
- Claude Code MCP docs: https://docs.claude.com/en/docs/claude-code/mcp

### Codex
Install command:
codex mcp add clearsky-ai --url https://staging-api.clearskyai.co.uk/mcp
Steps:
1. Run the command once in a terminal where `codex` is installed.
2. Start a Codex session and approve the browser sign-in flow when ClearSky AI needs authorization.
3. Use `codex mcp list` if you want to confirm the server is configured.
Docs:
- Codex MCP docs: https://developers.openai.com/learn/docs-mcp

## Browser connector installs
### Claude
Connector URL:
https://staging-api.clearskyai.co.uk/mcp
Steps:
1. Open Customize > Connectors, then choose Add custom connector.
2. Paste the ClearSky AI MCP URL, save it, and continue through the browser auth flow.
3. Enable the connector for the conversation where you want Claude to use ClearSky AI tools.
Note: Team and Enterprise workspaces need an owner to add the custom connector before members can connect it.
Docs:
- Claude connector docs: https://support.anthropic.com/en/articles/11175166-getting-started-with-custom-integrations-using-remote-mcp

### ChatGPT
Connector URL:
https://staging-api.clearskyai.co.uk/mcp
Steps:
1. Enable Developer Mode in ChatGPT under Settings > Connectors > Advanced.
2. Add a custom connector, paste the ClearSky AI MCP URL, and complete browser auth.
3. Use the connector from the Developer Mode tools picker when you want ChatGPT to call ClearSky AI.
Note: ClearSky AI exposes action-oriented MCP tools rather than the standard `search` and `fetch` pair, so ChatGPT currently needs Developer Mode for the full tool surface.
Docs:
- Developer Mode docs: https://platform.openai.com/docs/developer-mode
- MCP requirements: https://platform.openai.com/docs/mcp/

## Optional skill snippet
```text
Use ClearSky AI for real app work.
- Start with `search_tools` to find the right app operations or triggers.
- Call `get_tool_schemas` before executing unfamiliar tools or writing workflow code.
- Use `execute_tools` for direct actions, `run_code` for multi-step logic, and workflow tools only when the task should persist.
- If an app is not connected, use `connect_app` and wait for the browser handoff to finish.
```

## After setup
- Confirm the connector or MCP server is visible in the user's client before moving on.
- Finish the browser auth or consent flow whenever the client opens one.
- Ask what apps, actions, or workflows the user wants to run through ClearSky AI.