Axiom MCP Server – Observability Data for AI Agents
Axiom MCP Server lets AI agents query Axiom datasets, inspect schemas, work with metrics, and manage dashboards using the Model Context Protocol. It is useful when teams want an MCP-compatible assistant to investigate logs, events, metrics, monitors, and saved APL queries with Axiom permissions and authentication controls.
Overview
Axiom MCP Server is Axiom's official Model Context Protocol server for connecting AI agents to Axiom observability data. It gives MCP-compatible clients a structured way to query event data with Axiom Processing Language, inspect datasets, analyze metrics, and work with dashboards from the same environment where engineers ask operational questions.
What the MCP server enables
The server exposes tools for common Axiom investigation and reporting workflows. Agents can list datasets, read dataset schemas, execute APL queries with queryApl, run metric queries with queryMetrics, search metric names and tags, retrieve saved queries, inspect monitors and monitor history, and create, update, export, delete, or retrieve dashboards. Axiom also documents prompt templates for data-quality investigations, anomaly detection, baseline establishment, monitor-health analysis, unknown-dataset exploration, and correlation across datasets.
Axiom currently states that MCP resources are planned for future support, so this profile describes the documented tools and prompts rather than claiming resource support. The Axiom-hosted MCP server is located in the United States, and remote MCP query results are routed through US infrastructure.
When to use it
Use this MCP server when an AI agent needs to answer operational questions from live Axiom data, such as listing available datasets, reviewing the schema of a logs dataset, identifying common status codes in a recent time window, exploring metric tags, or turning dashboard JSON into a managed Axiom dashboard. It is especially useful for observability, incident investigation, product analytics, and AI-assisted data exploration where the agent should work from current Axiom data instead of static model knowledge.
Connection and authentication
Axiom supports a hosted remote MCP endpoint at https://mcp.axiom.co/mcp and an SSE endpoint at https://mcp.axiom.co/sse for clients that require server-sent events. Remote connections can authenticate with browser-based OAuth. For agents that cannot use browser-based OAuth, Axiom documents header-based authentication with Authorization: Bearer PERSONAL_ACCESS_TOKEN and x-axiom-org-id: ORGANIZATION_ID; when only the Authorization header is supported, the organization ID can be passed as an org-id URL parameter.
Axiom also supports a local self-run setup. The latest binary can be installed with go install github.com/axiomhq/axiom-mcp@latest; local authentication uses an Axiom API token, either in a config file or through environment variables such as AXIOM_TOKEN and AXIOM_URL.
Key considerations
Prefer the hosted remote endpoint for most users because Axiom documents it as the recommended approach and it supports OAuth credential isolation. Local setup offers more control but uses directly configured API tokens and does not support the documented MCP prompts. Store personal access tokens and API tokens outside client configuration files when possible, use least-privilege Axiom credentials, and revoke access from the Axiom profile sessions page when an AI client no longer needs access.
Supported Transports
streamable_http
URL: https://mcp.axiom.co/mcp
sse
URL: https://mcp.axiom.co/sse
stdio
Command: axiom-mcp
Frequently Asked Questions
- When should an AI agent use Axiom MCP Server?
- Use Axiom MCP Server when an AI agent needs to investigate live Axiom observability data, such as listing datasets, checking a dataset schema, querying recent logs with APL, exploring metrics, analyzing monitor history, or helping build and update dashboards.
- What does Axiom MCP Server add to an AI agent's capabilities?
- It gives the agent documented MCP tools for Axiom datasets, schemas, APL queries, metric queries, dashboards, saved queries, monitors, and monitor execution history, plus documented prompts for investigations such as anomaly detection and data-quality analysis.
- What can the AI agent access or manage through Axiom MCP Server?
- Depending on the authenticated user's permissions, the agent can query datasets, inspect schemas, search and query metrics, retrieve saved APL queries, list monitor configurations, inspect monitor history, and create, update, export, get, list, or delete Axiom dashboards. Axiom documents future support for MCP resources but does not currently list them as supported.
- How is authentication configured for Axiom MCP Server?
- The hosted remote server supports browser-based OAuth. Axiom also documents header-based authentication using an Authorization bearer token and an `x-axiom-org-id` header, or an `org-id` URL parameter when a client only supports the Authorization header. The local server uses an Axiom API token through a config file or environment variables such as `AXIOM_TOKEN`.
- Which transport should be used for Axiom MCP Server?
- Use `streamable_http` with `https://mcp.axiom.co/mcp` for remote MCP clients that support direct remote connections. Use the SSE endpoint only for agents that require server-sent events. Use the local `stdio` server when you need to self-run the server for more control, accepting that local setup uses API tokens directly and does not support Axiom's documented MCP prompts.