Quickstart

Before you begin

You will need a Queue account and an API key. Sign up at queue.dev, then go to Settings → API Keys and create a new key with the Agent Execution scope.

Step 1 — Install the CLI

Install the Queue CLI globally with npm. This gives you the queue command, which you will use to authenticate, run agents, and manage your workspace from the terminal.

npm

npm install -g @queue/cli

Homebrew

brew install queue-dev/tap/queue

Step 2 — Authenticate

Run queue login to open a browser-based sign-in flow. Your credentials are stored locally in ~/.queue/config.yaml. Alternatively, set the QUEUE_API_KEY environment variable directly if you prefer not to use the interactive flow.

Step 3 — Run your first agent

Navigate to any git repository on your machine and run the following command. Queue will spin up an agent, give it read access to your repo, and complete the task.

Example

queue run “Find all TODO comments in this repo and open a GitHub issue summarizing them”

The agent will read your files, gather the TODOs, format them into a structured summary, and — if you have the GitHub integration connected — open an issue automatically. You will see each step streamed to your terminal in real time.

Step 4 — Review the trace

After the run completes, open the Queue dashboard at queue.dev/runs or run queue trace <run-id> in your terminal. The trace shows every model call, tool invocation, file read or write, and decision the agent made.

This is one of the most important features in Queue. Every run is fully observable: you can audit what the agent did, replay individual steps, and understand exactly why it made the choices it did.

Step 5 — Connect your first integration

Head to Settings → Integrations and connect GitHub. Once connected, your agents can clone repos, read code, create branches, open pull requests, and post review comments — all from a single run.

What is next?

Now that you have a working agent, explore the Agents section to learn how to customize system prompts, add tools, and configure agent behavior for your specific workflow. Or jump to the SDK docs if you want to embed Queue agents into your own application.