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

# Use with Coding Agents

> Give AI coding agents HoneyHive skills, CLI access, and live documentation context.

If you're using an AI coding agent, give it four kinds of HoneyHive context:

<CardGroup cols={2}>
  <Card title="Agent Skills" icon="sparkles" href="#agent-skills">
    Reusable HoneyHive workflows your agent can follow.
  </Card>

  <Card title="HoneyHive CLI" icon="terminal" href="#honeyhive-cli">
    Terminal access to HoneyHive resources and API workflows.
  </Card>

  <Card title="Docs MCP Server" icon="plug" href="#docs-mcp-server">
    Live search across the HoneyHive docs from your agent.
  </Card>

  <Card title="llms.txt" icon="file-lines" href="#llmstxt">
    Paste a single URL for full doc context.
  </Card>
</CardGroup>

***

## Agent Skills

[Agent Skills](https://agentskills.io/) are portable `SKILL.md` packages that teach coding agents domain-specific workflows. HoneyHive publishes official skills at [github.com/honeyhiveai/skills](https://github.com/honeyhiveai/skills).

Use HoneyHive skills when you want an agent to perform a repeatable HoneyHive workflow, such as adding tracing to an LLM app, rather than just answering doc questions.

### Install HoneyHive skills

Install [GitHub CLI](https://cli.github.com/) v2.91.0 or later, which includes [`gh skill`](https://cli.github.com/manual/gh_skill), then install all recommended HoneyHive skills:

```bash theme={null}
gh skill install honeyhiveai/skills honeyhive-instrument
gh skill install honeyhiveai/skills honeyhive-evaluate
gh skill install honeyhiveai/skills honeyhive-improve
```

### Set up runtime prerequisites

These skills may ask your agent to run the HoneyHive CLI to validate credentials or inspect traces. Before using them, install the CLI from the [HoneyHive CLI docs](/v2/sdk-reference/cli), then set your API key:

```bash theme={null}
export HH_API_KEY="<your-project-api-key>"
```

For self-hosted, federated, or non-default deployments, set `HH_API_URL` to your deployment's API base URL. Most cloud users can omit it.

Validate access with a read-only CLI command:

```bash theme={null}
honeyhive datasets list
```

HoneyHive recommends installing all three skills together:

* `honeyhive-instrument` helps compatible agents add HoneyHive tracing to LLM, agent, and RAG applications.
* `honeyhive-evaluate` helps agents set up HoneyHive evaluations and experiments.
* `honeyhive-improve` helps agents use HoneyHive traces and evaluation results to improve application behavior.

<CardGroup cols={2}>
  <Card title="HoneyHive Skills" icon="github" href="https://github.com/honeyhiveai/skills">
    Browse available HoneyHive skills and installation notes.
  </Card>

  <Card title="Agent Skills Specification" icon="book" href="https://agentskills.io/specification">
    Learn how `SKILL.md` packages work across coding agents.
  </Card>
</CardGroup>

<Note>
  Install all three skills so your agent can instrument, evaluate, and improve HoneyHive-backed applications without switching setup flows.
</Note>

***

## HoneyHive CLI

The HoneyHive CLI gives coding agents a terminal-native way to inspect and manage HoneyHive resources. Use it when your agent needs to script datasets, runs, projects, or other REST API workflows from your development environment.

<CardGroup cols={2}>
  <Card title="HoneyHive CLI" icon="terminal" href="/v2/sdk-reference/cli">
    Install the CLI and run your first commands.
  </Card>

  <Card title="CLI Docs Home" icon="book" href="https://honeyhiveai.github.io/honeyhive-cli/">
    Browse the generated command reference.
  </Card>
</CardGroup>

For example, after setting `HH_API_KEY`, an agent can inspect recent trace events from the terminal:

```bash theme={null}
honeyhive events search --limit 10
```

Use the CLI with skills when the agent needs to validate credentials, inspect traces, or automate HoneyHive workflows as part of a code change.

***

## Docs MCP Server

HoneyHive documentation includes a built-in [Model Context Protocol (MCP)](https://modelcontextprotocol.io/) server. When connected, your AI assistant can search and retrieve HoneyHive docs in real time while generating responses, instead of relying on potentially outdated training data.

The HoneyHive docs MCP server is available at:

```txt theme={null}
https://docs.honeyhive.ai/mcp
```

The server exposes a `search_honey_hive_ai_docs` tool that performs semantic search across all HoneyHive documentation, returning relevant content with direct links to the source pages.

Once connected, you can ask your AI assistant questions about HoneyHive tracing, evaluations, integrations, and more. It searches the documentation directly to provide accurate, current answers.

<Note>
  The docs MCP server indexes both the current (v2) and legacy (v1) documentation. When prompting, tell your agent to use the **v2** docs so it pulls from the current SDK and UI references (see [Example Prompts](#example-prompts)).
</Note>

### Cursor

Open Cursor Settings with `Cmd + Shift + J` (Mac) or `Ctrl + Shift + J` (Windows/Linux), click **Tools & MCP** in the sidebar, then click **New MCP Server**. This opens `~/.cursor/mcp.json`. Add:

```json theme={null}
{
  "mcpServers": {
    "honeyhive-docs": {
      "url": "https://docs.honeyhive.ai/mcp"
    }
  }
}
```

Use `.cursor/mcp.json` in your project root instead if you want to commit the config and share it with your team.

### Claude Code

Run this command in your terminal to add the server to your current project:

```bash theme={null}
claude mcp add --transport http honeyhive-docs https://docs.honeyhive.ai/mcp
```

To make it available across all projects, add the `--scope user` flag:

```bash theme={null}
claude mcp add --transport http honeyhive-docs --scope user https://docs.honeyhive.ai/mcp
```

### VS Code / GitHub Copilot

Add the following to your VS Code MCP settings configuration file (`.vscode/mcp.json`):

```json theme={null}
{
  "servers": {
    "honeyhive-docs": {
      "type": "http",
      "url": "https://docs.honeyhive.ai/mcp"
    }
  }
}
```

Run `MCP: Open User Configuration` from the Command Palette to register the server across all workspaces instead of a single project.

### Windsurf

Add the following to your Windsurf MCP configuration:

```json theme={null}
{
  "mcpServers": {
    "honeyhive-docs": {
      "url": "https://docs.honeyhive.ai/mcp"
    }
  }
}
```

### Codex CLI

Run this command to register the server in `~/.codex/config.toml`:

```bash theme={null}
codex mcp add honeyhive-docs --url https://docs.honeyhive.ai/mcp
```

The `--url` flag requires Codex CLI with streamable HTTP support. Upgrade with `npm install -g @openai/codex` if `codex mcp add --url` is not recognized.

### Claude Desktop

Claude Desktop connects to remote MCP servers through custom connectors in the app, not through `claude_desktop_config.json` (which only supports local stdio servers).

1. Open Claude Desktop and go to **Settings** > **Connectors**.
2. Click **Add custom connector**.
3. Set the name to `honeyhive-docs` and the remote MCP server URL to `https://docs.honeyhive.ai/mcp`, then click **Add**.

On Team and Enterprise plans, an Owner must first add the connector under **Organization settings** > **Connectors** before members can enable it. See Anthropic's [custom connectors guide](https://support.anthropic.com/en/articles/11175166-using-custom-integrations-in-claude) for details.

### ChatGPT

ChatGPT connects to MCP servers through custom connectors:

1. Go to **Settings** > **Apps & Connectors** > **Advanced settings** and turn on **Developer mode**.
2. Return to **Settings** > **Apps & Connectors** and click **Create**.
3. Fill in a name (e.g. `HoneyHive docs`) and set the connector URL to `https://docs.honeyhive.ai/mcp`, then click **Create**.

Custom connectors with full MCP support are available on ChatGPT Business, Enterprise, and Education plans, and only workspace admins or owners can enable developer mode. See OpenAI's [connector setup guide](https://developers.openai.com/apps-sdk/deploy/connect-chatgpt) for details.

***

## llms.txt

The `llms.txt` file provides a structured overview of HoneyHive documentation optimized for LLM consumption. Include the URL in a prompt to give any AI assistant broad context about HoneyHive's capabilities.

```txt theme={null}
https://docs.honeyhive.ai/llms.txt
```

For the full documentation content in a single file:

```txt theme={null}
https://docs.honeyhive.ai/llms-full.txt
```

For example, you might prompt: *"Using the docs at [https://docs.honeyhive.ai/llms-full.txt](https://docs.honeyhive.ai/llms-full.txt), help me add OpenAI tracing to my Python project."*

***

## When to use Skills, CLI, MCP, or llms.txt

|                  | Agent Skills                           | HoneyHive CLI                    | MCP Server                              | llms.txt                               |
| ---------------- | -------------------------------------- | -------------------------------- | --------------------------------------- | -------------------------------------- |
| **Best for**     | Repeatable HoneyHive workflows         | Terminal actions and automation  | Ongoing development and IDE integration | One-off questions and quick context    |
| **How it works** | Loads reusable `SKILL.md` instructions | Runs `honeyhive` commands        | Searches live docs from the agent       | Provides a static docs index and links |
| **Setup**        | Install with `gh skill`                | Install CLI and set `HH_API_KEY` | Add server URL to your agent config     | Paste a URL                            |
| **Freshness**    | Versioned in GitHub                    | Uses installed CLI version       | Always live                             | Updated on each docs deployment        |

***

## Page Quick Actions

Every page in the HoneyHive docs has a contextual menu in the top-right corner with shortcuts for AI tools:

* **Copy as Markdown** - paste the full page content directly into any AI assistant
* **View as Markdown** - view the raw markdown source of the page
* **Open in ChatGPT** - open the page in ChatGPT with full context
* **Open in Claude** - open the page in Claude with full context
* **Connect MCP** - connect the docs MCP server to your tool
* **Add to Cursor** - add the docs MCP server directly to Cursor
* **Add to VS Code** - add the docs MCP server directly to VS Code

These actions are useful for quickly sharing context with your AI tool without a full MCP server setup.

***

## Example Prompts

Once you've connected HoneyHive to your agent, try these prompts. Scope docs searches to the current (`v2`) docs so the agent doesn't pick up pages from the legacy v1 tree.

<AccordionGroup>
  <Accordion title="Install HoneyHive tracing">
    ```
    Install the honeyhive-instrument skill from https://github.com/honeyhiveai/skills,
    then use it to instrument this project. Use the HoneyHive v2 docs for
    current SDK guidance.
    ```
  </Accordion>

  <Accordion title="Add OpenAI tracing to a Python project">
    ```
    Search the HoneyHive v2 docs (pages under /v2/) and add OpenAI
    tracing to my Python project using OpenInference. Install the
    required packages and initialize the HoneyHiveTracer with my
    API key.
    ```
  </Accordion>

  <Accordion title="Inspect HoneyHive from the terminal">
    ```
    Use the HoneyHive CLI docs and my HH_API_KEY environment variable to
    inspect recent events for this project.
    ```
  </Accordion>

  <Accordion title="Set up an experiment with evaluators">
    ```
    Using the HoneyHive v2 docs (pages under /v2/), help me set up
    an experiment that runs my RAG pipeline against a dataset and
    scores results with a custom Python evaluator.
    ```
  </Accordion>

  <Accordion title="Build a monitoring dashboard">
    ```
    Search the HoneyHive v2 docs (pages under /v2/) for how to
    create a monitoring dashboard that tracks latency, cost, and
    error rates for my production LLM application.
    ```
  </Accordion>
</AccordionGroup>