Hikari MCP Documentation

Endpoint

Use HTTP transport in MCP clients.

https://logs.example.com/mcp

Recommended workflow

Call get_instructions to load field mapping and investigation guidance.
Call summarize_window to inspect the current time range and major facets.
Use get_fields, get_field_values, or get_facets before writing narrow queries.
Use query_logs for known LogsQL. If AI is enabled, use ai_search for natural-language requests.
Keep result limits bounded and include the query in the final answer.

Optional AI capability

ai_search is registered only when Hikari has an OpenAI API key from OPENAI_API_KEY, or from the optional AWS Secrets Manager integration. Without a key, MCP clients still get the non-AI tools for raw queries, facets, field values, hit counts, summaries, and bounded live-tail samples.

Facet mapping

MCP summaries use the same configured facets as the UI. Configure those facets and aliases in config/field-mappings.json instead of hard-coding source fields into client prompts.

{
  "aliases": {
    "service": ["service", "service.name", "service_name", "kubernetes.container_name"],
    "host": ["host", "host.name", "host_name", "hostname", "kubernetes.pod_node_name"]
  },
  "facets": [
    { "field": "service", "label": "Service", "summary": true },
    { "field": "host", "label": "Host", "summary": true },
    { "field": "level", "label": "Level", "summary": true },
    { "field": "kubernetes.pod_namespace", "key": "namespace", "label": "Namespace", "summary": true },
    { "field": "kubernetes.pod_name", "key": "pod", "label": "Pod", "summary": true }
  ]
}

Hikari applies the aliases to backend VictoriaLogs requests with hidden LogsQL copy pipes. For example, host_name can populate the canonical host facet and service_name can populate service without changing the query shown to users or MCP clients.

Available tools

Tool Purpose

get_instructions Explains Hikari, field mapping, and the recommended workflow.

summarize_window Returns time buckets and UI-like facets for service, hostname, level, namespace, and pod.

query_logs Runs bounded VictoriaLogs LogsQL and returns rows plus count stats.

ai_search Optional. Generates LogsQL from a natural-language prompt when OpenAI is configured.

get_facets Returns grouped counts for selected fields.

get_fields Lists field names visible for a query window.

get_field_values Lists common values for one field within a query window.

get_hits Returns hit counts over time for a LogsQL query.

tail_logs Samples the live VictoriaLogs tail for a bounded duration or row count.

Example calls

get_instructions()

summarize_window({
  "query": "_time:15m",
  "step": "1m",
  "limit": 25
})

ai_search({
  "prompt": "summarize checkout errors from the last hour",
  "limit": 50
})

query_logs({
  "query": "_time:15m level:error service:\"api\"",
  "limit": 100
})

If ai_search is not returned by tools/list, configure OPENAI_API_KEY or use the non-AI query and discovery tools.