Add an App With Your Coding Agent

You can now ask your coding agent to set up AppSignal in your application. Give it a prompt, let it follow the installation instructions, and watch a checklist fill in as data starts arriving.

Installing AppSignal was already straightforward, but it still meant checking your environment, adding the right integration, creating an app, handling credentials, and confirming that telemetry arrived. These are useful checks, but they don't all need your attention. Your agent can work through them while you get on with something else.

We've been putting the pieces in place to make that possible: installation docs an agent can read, a provisioning API it can use, and a page that shows you how the setup is progressing. It's the first thing you do in AppSignal, and now it's something you can delegate.

Set up with AI was developed in AppSignal Labs and is now available to everyone as a Labs feature. It also uses some of the expert context we released earlier this week.

How It Works

The page walks you through three steps:

Start an agent in your project. Use any coding agent that can read a URL and edit files, including Claude Code, Cursor, or Codex. The instructions cover Ruby, Elixir, Node.js, Python, PHP, Go, Java, and JavaScript. Your agent follows the instructions for your project.

Paste the prompt. The page gives you a short prompt with a copy button. It includes the installation instructions and a provisioning URL for your organization:

Add AppSignal to this application.
 
Your job is to read AppSignal's installation instructions and follow them.
Retrieve them at https://docs.appsignal.com/agents/new-application.md
 
Your AppSignal provisioning status URL, which carries your provisioning token:
https://appsignal.com/api/provision/v1/status?token=prov-…

Watch the checklist. The page checks for progress while your agent works. It tracks four required steps and two optional logging steps:

• An app appeared in AppSignal
• Integration detected
• Telemetry arriving
• Requests traced
• Log source created (optional)
• Logs flowing (optional)

Once the four required steps are complete, the page takes you to your new app's dashboard.

What Makes This Possible

The prompt can be short because the installation instructions already live in our docs as plain Markdown. The agent starts with /agents/new-application.md, then follows the instructions for your application's language. Those docs have been available for a while, and you can still use them on their own.

The missing piece was provisioning. An agent could read the instructions, but you still had to create the app, copy a push API key into your project, and check whether data had arrived.

The provisioning URL connects those steps. It contains a short-lived token that lets the agent use an API to:

• Check what it can use: Confirm the token works and see the organization's existing apps, hosted collectors, and endpoints.
• Create or find the app: Resolve it by name and environment and receive its keys. Repeating the same request with the same body returns the same app instead of creating a duplicate.
• Provision a hosted collector if the integration needs one.
• Report the result: Record whether setup succeeded, partially succeeded, or failed, with notes.

You no longer need to copy keys between windows or create the app separately before configuring it.

What Can the Provisioning Token Access?

The URL contains a credential, so treat it as you would any token you give to a coding agent. Its permissions are limited to provisioning: it can access the information and actions needed for setup, but it can't read telemetry, change your settings, or touch billing.

The token has two expiry limits. Each accepted request moves its expiry forward by an hour, so an installation that keeps making progress can continue. It also has a fixed 24-hour limit from creation. Rejected requests don't extend either limit.

Tokens start with prov-, making them recognizable to secret scanners. If you want a new one, select Rotate this token on the page. That creates a fresh token and expires the old one.

You can always choose Set up by hand if you don't want to give a credential to an agent.

If Setup Takes Longer

An agent might take a while or pause to ask you a question. You can leave the page and come back to the checklist to see which steps have completed.

Before changing your code, the installation instructions run through five preflight checks to confirm the environment is ready. We shaped those instructions with practitioners who deploy Rails applications to production.

Keep Using AppSignal With Your Agent

Once setup is complete and your application is reporting, your agent can use its signals through AppSignal's context store. Connect it to the hosted MCP endpoint, which became generally available last week. It now has additional tools for check-in monitors, incident search, and deeper traces:

claude mcp add --transport http appsignal https://appsignal.com/api/mcp

Or give your agent the CLI if you want it to work in the same shell as your tests and deploys. AppSignal for agents explains when to use each option.

Set up with AI is available through Labs, and we'd like to hear where it gets stuck. Tell the team building it in our Discord. For step-by-step instructions, see Add a New Application With AI Agents.