Instructions for AI agents
These are reference documents written for LLMs and AI agents rather than people. They teach an agent how to work with Honeybadger correctly: writing BadgerQL, configuring visualizations and dashboards, managing alarms, and searching errors.
If you are building on Honeybadger with an AI assistant, coding agent, or your own tooling, fetch the relevant instructions and include them as context. The Honeybadger MCP server serves this same content to connected agents.
Available instructions
Section titled “Available instructions”| Name | Covers | Size |
|---|---|---|
| badgerql | The BadgerQL language: grammar, type hints, built-in fields, statements, expression functions, and the rules for writing correct queries. | ~7,593 tokens |
| queries | Fundamentals for querying Honeybadger Insights: streams, time ranges, event-class filtering, and verifying field names exist before aggregating. | ~1,524 tokens |
| charts | Visualization views for Insights query results and the chart_config fields each view accepts. | ~1,644 tokens |
| dashboards | Insights dashboard structure: the dashboard object, widget types and their configs, grid layout, and the vis object. | ~1,220 tokens |
| alarms | Insights alarms: alarm fields, trigger_config, states, evaluation timing, and query guidelines. | ~998 tokens |
| errors | The Honeybadger error model (faults and notices), lifecycle states, and the error search query language. | ~2,232 tokens |
Fetching them
Section titled “Fetching them”Each document is available in three forms:
- Raw text at
/resources/llms/instructions/<name>.txt. This is the document byte-for-byte, with no page formatting. Use this form when injecting instructions into an agent’s context. - A readable page at
/resources/llms/instructions/<name>/(linked from the table above), with a markdown version at/resources/llms/instructions/<name>.mdlike every page on this site. - A machine-readable catalog at /resources/llms/instructions/index.json, listing every document with its description, approximate token count, SHA-256 digest, and URL. Tools can read the catalog to discover what’s available without hardcoding the list, and use the digest to skip re-downloading unchanged content.