> ## Documentation Index
> Fetch the complete documentation index at: https://docs.appsignal.com/llms.txt
> Use this file to discover all available pages before exploring further.

# AppSignal for agents

> Give AI agents and coding assistants access to your AppSignal data through the MCP server or the CLI.

Your monitoring data is most useful where you're already working. AppSignal MCP connects AI agents and editors directly to your data, and the AppSignal CLI puts the same data in your terminal, with machine-readable output for agents to consume.

<CardGroup cols={2}>
  <Card title="AppSignal MCP" icon="plug" href="/mcp-server">
    Connect Claude, Cursor, Windsurf, Zed, VS Code, or GitHub Copilot to your errors, traces, logs, and metrics.
  </Card>

  <Card title="AppSignal CLI" icon="terminal" href="/cli">
    Tail logs, inspect incidents, and manage triggers from your shell, in human-readable or JSON output.
  </Card>
</CardGroup>

<h2 id="which-one-to-use">
  Which one to use
</h2>

Both reach the same data. They differ in how the agent talks to AppSignal.

* **MCP** is a hosted HTTP endpoint at `https://appsignal.com/api/mcp`. Point your editor or chat app at it once and the agent queries AppSignal directly, with no install step. Read the [setup instructions](/mcp/setup) for your tool and the [tool reference](/mcp/reference) for what it exposes.
* **The CLI** is a single binary you run yourself. Agents drive it by running commands and reading `--output json`. Use it when you want the agent working in the same shell as your tests and deploys, or when the environment restricts outbound traffic.

<h2 id="teach-your-agent-the-cli">
  Teach your agent the CLI
</h2>

The CLI bundles an agent skill: a set of instructions that teaches Claude, Codex, or OpenCode when to reach for `appsignal-cli` and how to read its output.

```sh Shell theme={null}
appsignal-cli skill install
```

Read the [agent skill](/cli/skill) page for what it installs and how to keep it current.

<h2 id="what-agents-can-reach">
  What agents can reach
</h2>

Both surfaces read the data AppSignal already has. If your application isn't sending logs, an agent can't query them.

* Errors and exception incidents
* Performance incidents and traces
* Logs, including saved views and log-based metrics
* Metrics and dashboards
* Anomaly detection triggers
* Applications, namespaces, and deploy markers

<Note>
  An agent using MCP or the CLI acts as you, with your access to your
  organization's data. Authenticate the [CLI](/cli/authentication) on the machine
  where the agent runs, and treat MCP access the same way you treat your own.
</Note>

<h2 id="building-your-own-integration">
  Building your own integration
</h2>

If you want to query or push data from your own code rather than from an agent, use the [AppSignal API](/api) instead.

<h2 id="documentation-an-agent-can-read">
  Documentation an agent can read
</h2>

Every page here is served as plain Markdown at its URL with a `.md` extension, so an agent can read the documentation as well as the data.

The [Ruby and Rails tutorials](/tutorials), produced with GoRails, are written for both audiences: framework walkthroughs with requirements in tables, steps in order, and an explicit result for each step. They document Rails rather than AppSignal, which is the context an agent reading a metric from a Rails application does not otherwise have.

If you want your own application to answer in Markdown, the [Markdown renderer tutorial](/tutorials/ruby/markdown-renderer) covers the `text/markdown` MIME type Rails 8.1 added.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.