Flushnet Help

Start here for new users

1. Open the chat first: open AI Runner Chat in ChatGPT.
2. Log in to Flushnet: use Login or create an account with Create Account.
3. Open dashboard: go to Dashboard after login.
4. Generate your access code: open Generate Access Code and copy the code.
5. Paste the code in chat: return to the AI Runner chat and paste the code when requested.
6. Run your first task: ask the chat to list nodes, use sandbox, or run a simple safe command.

Open AI Runner Chat · Generate Access Code · Install runtime · Support

Beginner guide

What Flushnet does: it lets you run actions on your own computer or in a safe sandbox from a web page or a AI Runner chat.

• Native mode: controls your real computer.
• Sandbox mode: runs commands inside an isolated browser Linux-like environment.
• Dashboard: human-friendly control panel.
• GPT actions: lets ChatGPT control your nodes through your API.

Basic flow: create account → log in → generate access code → open dashboard or GPT → choose node → run task.

What is a node? A node is a machine or environment connected to Flushnet. Examples: your Mac, a Windows PC, a sandbox, or a GPU server.

Step-by-step: use the AI Runner chat

1. Register an account: open Create Account and create your Flushnet account.
2. Log in: open Login and sign in to your Flushnet account.
3. Check your plan: open Subscription and confirm your account is active. Pro and Power plans support non-expiring chat access codes.
4. Generate an access code: open Generate Access Code and copy the code shown on screen.
5. Open the AI Runner chat: open AI Any Computer / Desktop Operator.
6. Provide the access code in chat: paste the Flushnet access code into the GPT chat when asked so the GPT can link the session to your account.
7. Grant access to your nodes: once the code is accepted, ask the GPT to list your nodes, then choose the machine or sandbox you want to control.
8. Run a simple test first: ask the GPT to run a safe command such as pwd, dir, or echo hello before trying more advanced tasks.
9. Use sandbox first when possible: for safer testing, start with a sandbox or WebVM node before using a native machine.

Install tools on your local computer

1. Open the installer page: go to Installer or download the runtime bundle from the website.
2. Install the runtime: on Windows or macOS, extract or run the provided launcher/runtime bundle.
3. Find your owner information: open the dashboard or Subscription page and note your Owner ID.
4. Use the runtime config: install the runtime with your node id, owner id, and token so the machine registers under your account.
5. Start the runtime: launch the runtime so it starts the local agent and related services on your computer.
6. Verify node registration: return to the Dashboard and confirm that your machine appears in the node list as online.
7. Test from the dashboard first: run one simple command from the dashboard to confirm the local agent works before using the AI Runner chat.
8. Then use the AI Runner chat: after the node is online, open the AI Runner chat, paste your access code, and ask it to use that machine.

Recommended order: register → log in → install runtime on local computer → verify node online in dashboard → generate access code → open AI Runner chat → paste code → run a test command.

Example use cases and workflows

Choose the workflow that fits your level and goal. Each example shows a practical step-by-step path.

Beginner hobbyists

1) Run your first safe command in sandbox

1. Create your Flushnet account and log in.
2. Open Dashboard.
3. Open the sandbox or choose a sandbox node.
4. Enter a simple command such as pwd, ls, or echo hello.
5. Run the command and read the output.
6. Repeat with one more safe command so you get comfortable with the flow.
7. Only move to native machine execution after the sandbox workflow feels clear.

2) Ask the AI Runner chat to use your sandbox

1. Log in to Flushnet.
2. Generate an access code from Generate Access Code.
3. Open the AI Runner chat here: Open the Flushnet ChatGPT agent, then paste the access code.
4. Ask the GPT to list your nodes.
5. Tell it to use your sandbox node.
6. Ask it to run a simple command such as echo hello from sandbox.
7. Read the result and confirm the GPT is connected to the correct environment.

3) Install one local computer and test it

1. Open Installer.
2. Install the runtime on your Windows or macOS computer.
3. Start the runtime and wait for the machine to register.
4. Open the dashboard and confirm the node appears online.
5. Run one harmless test command such as dir on Windows or pwd on macOS.
6. Check the output in the dashboard before trying more advanced automation.

Professional users

1) Browser automation on a Windows node

1. Install the local runtime on a Windows machine.
2. Verify the Windows node is online in the dashboard.
3. Generate an access code and link the AI Runner chat if you want assistant-driven control.
4. Choose the Windows node from dashboard or GPT.
5. Start with a simple browser action such as opening a URL.
6. Then add a second action like click, type, or screenshot.
7. Review the result and keep the workflow as a repeatable operator task.

2) Test in sandbox, then promote to native

1. Write the command or workflow you want to use.
2. Run it first in sandbox to check behavior and output.
3. Adjust the command if the result is not what you expect.
4. When the sandbox result looks correct, choose your native node.
5. Run the same command or a production-safe version on the native machine.
6. Compare outputs and confirm the native run succeeded.

3) Run the same task across more than one machine

1. Register two or more machines under the same account.
2. Confirm all nodes show as online.
3. Choose a common task such as checking disk space, listing a directory, or running a script.
4. Submit the task to each node.
5. Watch status and results from the dashboard.
6. Compare outputs and identify failures, slow nodes, or differences between machines.

Enterprise teams

1) Standardized rollout across a team of machines

1. Define the machines that will become Flushnet nodes.
2. Install the runtime on each approved machine.
3. Verify that each node registers under the intended account or shared control model.
4. Use a standard verification command on every machine.
5. Document the results so you know which machines are ready for production workflows.
6. Move on to browser, file, or job workflows only after the baseline rollout is clean.

2) Operations workflow with approval path

1. Test a new command or automation flow in sandbox first.
2. Review output and confirm it is safe for production.
3. Run the validated version on a staging native machine.
4. Check logs, file effects, and output.
5. Only then apply the same workflow to the rest of the approved nodes.
6. Track each run through dashboard history or job status checks.

3) Assistant-driven execution with internal controls

1. Create the internal rule for who can generate and use access codes.
2. Choose which nodes can be controlled through the AI Runner chat or API path.
3. Generate the access code from a logged-in admin or operator account.
4. Link the assistant session to that account.
5. Start with read-only or harmless validation tasks such as listing nodes or reading status.
6. Move to controlled execution tasks only after the session, nodes, and outputs are verified.
7. Rotate or regenerate access when the session changes or the work period ends.

Professional overview

Flushnet is a brokered remote execution platform. Clients authenticate with X-API-Key, end users authenticate with a temporary access_code, nodes authenticate with their token. The broker maps commands to nodes, nodes poll for work, execute, and push results back. Access control supports private, shared, and public nodes.

• Control plane: PHP API, broker, auth, billing, access control.
• Execution plane: native node agent, sandbox agent, optional GPU/AI node.
• Interfaces: dashboard, AI Runner chat, installers, future live view/terminal.

GitHub integration quickstart

1. Create or choose a GitHub repository with a workflow that supports workflow_dispatch.
2. Get a GitHub token for the repository. For quick testing you can use a personal access token; for production use a GitHub App installation token.
3. Call /tools/github_list_workflows.php to confirm the repository workflows are visible.
4. Call /tools/github_dispatch.php with github_owner, github_repo, workflow_id, ref, and any inputs.
5. Call /tools/github_run_status.php with the run id to track status and conclusion.
6. Use this path for CI jobs, browser tests, repo automation, deployment workflows, or build pipelines that belong in GitHub Actions instead of a live node.

Service descriptions

• Dashboard: manual control UI for nodes, commands, files, and help pages.
• Native Agent: runs on a real user machine and exposes desktop commander / browser automation services through the broker.
• Sandbox: browser-contained Linux-like environment for safe trials, tutorials, and code execution without touching the real machine.
• GPT Actions API: lets a AI Runner chat list nodes, run commands, write files, and later submit jobs or AI tasks.
• File Tools: upload, download, read, write, and list directory operations.
• Terminal: interactive shell session for live command execution.
• Browser Automation / Live View: screenshots or future live streaming from browser/desktop sessions.
• Jobs: queued or parallel command execution across multiple nodes.
• AI / GPU Nodes: specialized nodes for inference, transcription, Python/CUDA jobs, and model execution.
• Billing & Subscription: plan-based access, usage tracking, quotas, and PayPal activation.

Safety notes

• Use sandbox when you want a safer environment.
• Use native only when you want to act on your real machine.
• Do not share access codes publicly.
• Only install agents on machines you own or are authorized to manage.

Smart device commands

If your Tuya device is linked, AI Runner can control it from chat.

• power on the smart plug
• power off the smart plug
• status of the smart plug

Current live alias: smart plug maps to Smart Socket.