---
title: OpenClaw | Lightcone
description: Give your OpenClaw AI assistant cloud computers powered by Lightcone.
---

Use Lightcone with [OpenClaw](https://openclaw.ai) to give your AI assistant cloud computers that Northstar can see and operate. Ask your assistant to browse the web, fill forms, extract data, and automate multi-step workflows — from WhatsApp, Telegram, Discord, or any connected channel.

## Install

Terminal window

```
openclaw plugins install @tzafon/openclaw-lightcone
openclaw plugins enable lightcone
openclaw config set env.TZAFON_API_KEY "sk_..."
```

Allow the Lightcone tools for your agent:

Terminal window

```
openclaw config set tools.allow '["lightcone_browse", "lightcone_session_create", "lightcone_session_action", "lightcone_session_close"]'
```

Restart the gateway to load the plugin.

## Example

Once installed, message your OpenClaw agent:

> “Go to news.ycombinator.com and tell me the top 3 stories”

The agent calls `lightcone_browse`, which creates a cloud computer, has Northstar navigate to the page, read the screen, and return the content with a screenshot.

## Tools

The plugin provides four tools at two levels of abstraction:

**High-level** — describe what you want done and Northstar operates a cloud computer to complete it:

- `lightcone_browse` — give it a natural language instruction and an optional starting URL

**Low-level** — create a cloud computer and send actions directly:

- `lightcone_session_create` — create a cloud computer (browser or desktop)
- `lightcone_session_action` — send actions: click, type, scroll, navigate, screenshot
- `lightcone_session_close` — shut down the computer

The plugin also ships two skills (`lightcone-browse` and `lightcone-session`) that teach the agent when and how to use these tools.

## How it works

The `lightcone_browse` tool runs a full [computer-use loop](/guides/cua-protocol/index.md) under the hood:

1. Creates a Lightcone cloud computer
2. Takes a screenshot and sends it to Northstar via the [Responses API](/guides/responses-api/index.md)
3. Northstar sees the screen and decides what to do — click, type, scroll, navigate
4. The plugin executes the action and feeds the next screenshot back
5. Repeats until Northstar reports the task is done
6. Returns the final page content and a screenshot to the OpenClaw agent

The session tools (`lightcone_session_*`) give you direct control over each step instead.

## Resources

- [Plugin source on GitHub](https://github.com/tzafon/openclaw-lightcone)
- [Plugin on npm](https://www.npmjs.com/package/@tzafon/openclaw-lightcone)
- [OpenClaw documentation](https://docs.openclaw.ai)

## See also

- [**Run a task**](/guides/run-a-task/index.md) — use Lightcone’s Tasks API directly
- [**Computers**](/guides/computers/index.md) — create and control cloud computers via the API
- [**MCP Server**](/integrations/mcp/index.md) — expose Lightcone as tools via Model Context Protocol