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

# Bronto Agent Skills

> Install Bronto's agent skills and plugins for Cursor or Claude Code so coding agents can investigate telemetry and improve logging workflows.

Bronto Agent Skills is a GitHub repository that packages Bronto-specific instructions and plugin metadata for AI coding assistants.

Use it when you want an agent to understand Bronto's MCP tools, investigate logs and traces from your editor, or run Bronto's Claude Code logging workflow.

<Note>
  The Cursor plugin and Claude Code plugin currently solve different jobs. The
  Cursor plugin installs the Bronto MCP configuration and Bronto investigation
  skill. The Claude Code marketplace plugin currently provides the
  `bronto-logging` workflow for adding stable statement IDs to logging calls.
</Note>

Repository:

```text theme={"dark"}
https://github.com/brontoio/bronto-agent-skills
```

## What's included

<CardGroup cols={2}>
  <Card title="Bronto investigation skill" icon="magnifying-glass">
    Teaches an agent how to use Bronto MCP for dataset discovery, schema
    inspection, timeseries analysis, raw log search, traces, latency, and
    production debugging.
  </Card>

  <Card title="Cursor plugin" icon="terminal">
    Bundles the Bronto hosted MCP server configuration for US and EU regions
    with the Bronto investigation skill.
  </Card>

  <Card title="Claude Code plugin marketplace" icon="plug">
    Adds the `bronto-logging` Claude Code plugin, including a statement ID
    supervisor skill and worker agent for updating logging calls across a
    codebase.
  </Card>

  <Card title="MCP configuration" icon="server">
    Defines hosted MCP endpoints for `bronto-us` and `bronto-eu`. OAuth is the
    default, with regional API keys available as an optional alternative.
  </Card>
</CardGroup>

## Before you start

You need:

* A Bronto account in the correct data region, US or EU
* MCP access enabled by a Bronto administrator under **Settings → Authentication**
* Optionally, a Bronto API key if you prefer API-key authentication to OAuth
* Cursor or Claude Code, depending on the workflow you are setting up

If you choose API-key authentication, create or manage keys in [API Keys](../Account-Management/API-Keys).

Use the MCP endpoint that matches your Bronto organization:

| Region | Hosted MCP endpoint            |
| ------ | ------------------------------ |
| US     | `https://mcp.us.bronto.io/mcp` |
| EU     | `https://mcp.eu.bronto.io/mcp` |

## Use with Cursor

The Cursor plugin is the most direct way to use this repository for Bronto telemetry investigation from an IDE. It installs:

* `mcp.json` with `bronto-us` and `bronto-eu` hosted MCP servers
* `skills/bronto/SKILL.md`, which tells the agent how to investigate Bronto telemetry
* Reference material for query shapes, MCP tool order, and investigation patterns

The plugin is also listed in the [Cursor Plugin Directory](https://cursor.directory/plugins/bronto). See the [Cursor plugin README](https://github.com/brontoio/bronto-agent-skills/tree/main/.cursor-plugin) for the latest installation and configuration details.

<Steps>
  <Step title="Install the plugin">
    Start the Cursor Agent CLI with `agent`, run:

    ```text theme={"dark"}
    /plugin
    ```

    Paste `https://github.com/brontoio/bronto-agent-skills.git` into plugin
    search, then install it at user or project scope.
  </Step>

  <Step title="Sign in with OAuth">
    Open **Customize → MCPs**, select `bronto-us` or `bronto-eu` for your
    organization's region, then select **Login** and complete the browser
    authentication flow. Disable the unused regional server.
  </Step>

  <Step title="Optionally configure an API key">
    If you do not want to use OAuth, open the installed Bronto plugin in
    **Customize**, select **Configure**, and set `US_BRONTO_API_KEY` or
    `EU_BRONTO_API_KEY` for your organization's region.
  </Step>

  <Step title="Reload Cursor">
    Run **Developer: Reload Window** from the command palette.
  </Step>

  <Step title="Check the MCP connection">
    Open Cursor's MCP status and confirm that either `bronto-us` or `bronto-eu`
    is connected. Disable or ignore the region you do not use.
  </Step>

  <Step title="Ask Cursor to investigate with Bronto">
    Try a prompt like:

    ```text theme={"dark"}
    Use Bronto to list the datasets available in this organization, then show the keys for the production API logs.
    ```
  </Step>
</Steps>

## Use with Claude Code

There are two Claude Code paths, depending on what you want to do.

### Investigate Bronto telemetry

For telemetry investigation in Claude Code, connect Claude directly to Bronto's hosted MCP endpoint.

```bash theme={"dark"}
claude mcp add --transport http bronto https://mcp.eu.bronto.io/mcp
```

Use the US endpoint instead if your Bronto account is hosted in the US:

```bash theme={"dark"}
claude mcp add --transport http bronto https://mcp.us.bronto.io/mcp
```

Then run:

```text theme={"dark"}
/mcp
```

Select Bronto and complete the authentication flow. See [Hosted MCP Server](./hosted-mcp) for the full setup guide.

### Use the Claude Code logging plugin

The Claude Code marketplace in `bronto-agent-skills` exposes the `bronto-logging` plugin. Use it when you want Claude Code to add or update stable `stmt_id` values on logging statements across a codebase.

<Steps>
  <Step title="Open the Claude Code plugin menu">
    ```text theme={"dark"}
    /plugin
    ```
  </Step>

  <Step title="Add the Bronto marketplace">
    Choose **Add Marketplace**, then enter:

    ```text theme={"dark"}
    brontoio/bronto-agent-skills
    ```
  </Step>

  <Step title="Enable the logging plugin">
    Go to **Manage and uninstall plugins**, select
    **bronto-claude-code-plugins**, then enable **bronto-logging**.
  </Step>

  <Step title="Run the statement ID workflow">
    Ask Claude Code to add statement IDs to logging calls in a directory:

    ```text theme={"dark"}
    Use the Bronto logging plugin to add stable statement IDs to logging calls under src/.
    ```
  </Step>
</Steps>

The plugin discovers logging calls, dispatches worker agents by file, and merges the generated mapping into `statementIds.json`.

## How the Bronto investigation skill works

The Bronto skill gives agents a repeatable query workflow:

1. Discover datasets with `get_datasets` or `get_datasets_by_name`.
2. Inspect fields with `get_keys`.
3. Inspect useful dimensions with `get_key_values`.
4. Use `timeseries` first for counts, rates, percentiles, group-bys, regressions, and top offenders.
5. Use `search_logs` for representative raw examples, exact trace IDs, event IDs, pasted errors, and payload drill-down.

This prevents the agent from guessing field names, querying every dataset, or jumping straight to raw logs before it has an aggregate view.

## Verify the setup

After installing either integration, ask the agent:

```text theme={"dark"}
List the Bronto datasets I can access, then show the available keys for one production dataset.
```

A working setup should return dataset names, collections, dataset IDs, and field names from the correct Bronto region.

If authentication fails, check:

* You are using the endpoint for the correct Bronto region
* The matching regional API key is configured if using API-key auth
* MCP login is enabled in Bronto if using OAuth
* Cursor third-party plugins, skills, and configs are enabled if using Cursor