> ## Documentation Index > Fetch the complete documentation index at: https://docs.honeycomb.io/llms.txt > Use this file to discover all available pages before exploring further. # Agent Skills for AI Coding Assistants > Give your AI coding assistants deep knowledge of observability, OpenTelemetry instrumentation, production debugging, and Honeycomb features. [Honeycomb Agent Skills](https://github.com/honeycombio/agent-skill) is a collection of skills, agents, and hooks for enhancing AI-assisted code workflows. These skills provide OpenTelemetry best practices, Honeycomb feature usage, production issue debugging, and more. ## Available skills Installing the Honeycomb plugin gives your AI assistant eight skills for observability workflows: * **query-patterns**: Build effective Honeycomb queries with patterns for column selection, filtering, and aggregation. * **production-investigation**: Debug latency spikes, error rates, and other production anomalies with step-by-step reasoning. * **slos-and-triggers**: Define and manage SLOs and alert triggers in Honeycomb. * **otel-instrumentation**: Instrument applications with OpenTelemetry, including span creation, attribute naming, and context propagation. * **otel-migration**: How to migrate proprietary or legacy instrumentation to OpenTelemetry. * **beeline-migration**: Guidlines for migrating from Honeycomb Beelines to OpenTelemetry SDKs. * **observability-fundamentals**: Core concepts around distributed tracing, wide events, and production-grade observability. * **create-honeycomb-board**: Create Honeycomb boards to capture investigations and share insights. ## Available agents Two autonomous agents are included for more complex tasks: * **honeycomb-investigator**: Autonomously investigates production issues by querying Honeycomb data, identifying patterns, and summarizing findings. * **instrumentation-advisor**: Analyzes your codebase to find instrumentation gaps and suggest improvements based on OpenTelemetry best practices. ## Additional capabilities * **Column validation hooks**: Prevent queries with invalid or mistyped column names before they are sent to Honeycomb. * **Schema caching**: Reduce repeated lookups by caching dataset column schemas locally during a session. * [Honeycomb MCP](/integrations/mcp/) integration: Query traces, check SLOs and triggers, run BubbleUp analysis, and create boards. ## Install the plugin The following tools support direct plugin installation, which sets up all skills, agents, hooks, and MCP configuration. ### Claude Code ```bash theme={} claude plugin marketplace add honeycombio/agent-skill claude plugin install honeycomb ``` After installation, run `/honeycomb-setup` to connect the Honeycomb MCP server and complete authentication. ### Cursor Add as a remote rule in the Cursor settings: 1. Open **Settings** > **Rules** > **Project Rules**. 2. Select **Add Rule** > **Remote Rule**. 3. Enter the URL: `https://github.com/honeycombio/agent-skill` ### Augment (Auggie CLI) ```bash theme={} auggie plugin marketplace add honeycombio/agent-skill auggie plugin install honeycomb ``` ### GitHub Copilot CLI ```bash theme={} copilot plugin install honeycombio/agent-skill:honeycomb ``` ### Other tools Many AI coding tools support skills and MCP servers through their own configuration. To use Honeycomb with these tools, you'll generally need to: 1. **Add skills**: Copy the skill files from the [agent-skill repository](https://github.com/honeycombio/agent-skill) into the skills directory your tool expects. 2. **Configure the MCP server**: Point your tool at the Honeycomb MCP endpoint — `https://mcp.honeycomb.io/mcp` (or `https://mcp.eu1.honeycomb.io/mcp` for EU). Refer to your tool's documentation for specifics: | Tool | Skills Docs | MCP Docs | | --------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | -------------------------------------------------------------------------------- | | VS Code with GitHub Copilot | [Agent Skills](https://code.visualstudio.com/docs/copilot/customization/agent-skills), [Agent Plugins](https://code.visualstudio.com/docs/copilot/customization/agent-plugins) | [MCP Servers](https://code.visualstudio.com/docs/copilot/chat/mcp-servers) | | OpenAI Codex CLI | [Skills](https://developers.openai.com/codex/skills) | [Configuration](https://developers.openai.com/codex/config-reference) | | Cline | [Rules](https://docs.cline.bot/features/cline-rules) | [MCP Servers](https://docs.cline.bot/mcp/mcp-overview) | | Windsurf | — | [MCP Configuration](https://docs.windsurf.com/windsurf/mcp) | | Amazon Q Developer | — | [MCP Servers](https://docs.aws.amazon.com/amazonq/latest/qdeveloper-ug/mcp.html) | For tools that only support MCP (not skills), your agent will have access to Honeycomb query tools but won't have the built-in observability reasoning and instrumentation guidance that skills provide. See the [Honeycomb MCP configuration guide](/integrations/mcp/configuration-guide/) for more details on connecting to the MCP server. Don't see your tool listed? [Open an issue on GitHub](https://github.com/honeycombio/agent-skill/issues) to request support. ## Connecting to Honeycomb ### OAuth (recommended) OAuth is the recommended authentication method. It uses a browser-based flow and does not require you to manage API keys. When you first invoke a Honeycomb MCP tool, your agent will open a browser window to complete the authorization flow. Pick your Honeycomb environment and grant the required permissions. Once authorized, your agent can access Honeycomb data for the rest of the session. ### API Key (headless environments) For autonomous or unattended agents that cannot complete an interactive OAuth flow, use an API key. 1. Generate a Management API key in Honeycomb under **Account** > **Team Settings** > **API Keys**. 2. Grant the **Model Context Protocol** and **Environments** scopes with **Read** permissions (and **Write** if you want the `create_board` tool). 3. Format the key as `KEY_ID:SECRET_KEY` and pass it as a Bearer token in the `Authorization` header. See the [configuration guide](/integrations/mcp/configuration-guide/#setting-up-an-api-key) for a full example configuration. ### EU region If your Honeycomb team is on the EU instance, use `https://mcp.eu1.honeycomb.io/mcp` in place of the US endpoint in all configurations above. ## Next steps * [Getting Started with Honeycomb Using MCP](https://academy.honeycomb.io/app/courses/13ad6a1c-2cd9-46d3-b6c3-c7dc45b5bf4f) * [Connect your AI agent to Honeycomb MCP](/integrations/mcp/configuration-guide/) * Browse the [honeycomb/agent-skill repo](https://github.com/honeycombio/agent-skill), open issues, or contribute new skills and improvements.