Axiom MCP Server: Natural-Language Logs, Traces, Events

Overview

The Axiom MCP Server implements the Model Context Protocol (MCP) for Axiom, letting AI agents query logs, traces, and events using natural-language interfaces. It maps MCP tool calls to Axiom Processing Language (APL) operations so models can run searches, inspect schemas, and fetch saved queries or monitor history without manually constructing API requests.

Note: the original repository for this implementation is deprecated and no longer maintained. For the maintained, official Axiom MCP Server, see https://mcp.axiom.co. The deprecated source is available at https://github.com/axiomhq/mcp-server-axiom for reference and self-hosted experimentation only.

This guide explains what the server provides, how to install and configure it, the MCP tools it exposes, and practical examples for troubleshooting and analysis workflows.

Features

• Natural-language friendly: exposes Axiom datasets and queries through MCP tools so LLMs can reason over log and trace data.
• APL execution: run arbitrary APL queries against Axiom datasets.
• Dataset discovery: list datasets and retrieve dataset schemas programmatically.
• Saved queries and monitoring: enumerate saved/starred APL queries and inspect monitor configurations and execution history.
• Rate limiting controls: configurable rate and burst limits for query and listing operations.
• Simple auth: uses an Axiom API token for authentication.

Installation / Configuration

Two common install options: download a prebuilt binary (releases) or install from source with Go.

Install from releases

# download latest built binary from the releases page:
# https://github.com/axiomhq/mcp-server-axiom/releases
# unpack and place the binary in your PATH
./axiom-mcp --help

Install from source (Go)

go install github.com/axiomhq/axiom-mcp@latest
# binary will be installed to $GOPATH/bin or $GOBIN

Configuration file example (config.txt)

# Axiom API token
token xaat-your-api-token

# Axiom API base URL (defaults to https://api.axiom.co)
url https://api.axiom.co

# Rate limits (queries per second) and bursts
query-rate 1
query-burst 1
datasets-rate 1
datasets-burst 1
monitors-rate 1
monitors-burst 1

Command-line flags

axiom-mcp \
  -token xaat-your-api-token \
  -url https://api.axiom.co \
  -query-rate 1 \
  -query-burst 1

Environment variables

export AXIOM_TOKEN=xaat-your-api-token
export AXIOM_URL=https://api.axiom.co
export AXIOM_QUERY_RATE=1
export AXIOM_QUERY_BURST=1

Integrating with an MCP client (example: Claude desktop)

{
  "mcpServers": {
    "axiom": {
      "command": "/path/to/axiom-mcp",
      "args": ["--config", "/path/to/config.txt"],
      "env": {
        "AXIOM_TOKEN": "xaat-your-api-token",
        "AXIOM_URL": "https://api.axiom.co"
      }
    }
  }
}

Available Tools

This implementation exposes MCP tools (not resources or prompts). Each tool requires an API token for authentication.

• queryApl — Execute an APL query against one or more Axiom datasets and return results (used for ad-hoc analytics and search).
• listDatasets — Return available Axiom datasets (used for discovery and autosuggest).
• getDatasetSchema — Fetch the schema for a target dataset (fields, types, metadata).
• getSavedQueries — List saved / starred APL queries accessible to the token (useful for reusing known investigations).
• getMonitors — List monitoring rules and configurations.
• getMonitorsHistory — Retrieve execution history for monitors (alerts, last runs, status).

Tool behavior notes:

• Each tool is rate-limited by the server settings.
• queryApl runs arbitrary APL; ensure the provided token has appropriate permissions.

Use Cases

1. Quick failure triage (natural language)

   • Ask your MCP-capable assistant: “Show the top 10 services with the highest 5xx error rate in the last 30 minutes and include sample traces.”
   • The assistant uses listDatasets/getDatasetSchema to find the right logs table, then queryApl to run an APL aggregation and surface results.

2. Investigate a saved diagnostic query

   • Use getSavedQueries to retrieve a bookmarked APL investigation, then run queryApl to execute and iterate on results interactively.

3. Monitor health and recent alerts

   • Use getMonitors to list active monitors and getMonitorsHistory to fetch recent trigger events. Useful for asking: “Which monitors fired in the last 24 hours and which hosts were affected?”

4. Build an AI-driven runbook

   • Combine dataset schema discovery with queryApl to let an AI produce step-by-step diagnostics: e.g., “If latency > 500ms on endpoint X, run these APL queries and recommend remediation steps.”

Example APL snippet (run via queryApl)

from logs
| where service == "payments" and status >= 500
| summarize count = count() by bin(1m), host
| sort by count desc
| limit 10

Troubleshooting & Notes

• Authentication: always provide a valid Axiom token (xaat-…). Tokens govern which datasets, saved queries and monitors are visible.
• Deprecation: this specific repository is deprecated. For production use or maintained integration, prefer the official Axiom MCP Server at https://mcp.axiom.co.
• Permissions: ensure the token has read access to datasets and monitors you want the MCP server to expose.
• Security: run the MCP server within trusted infrastructure; the server executes queries against your telemetry data.

For the code and legacy reference, see the archived GitHub repo: https://github.com/axiomhq/mcp-server-axiom