Chronosphere MCP Server

The Chronosphere MCP server is the official integration from Chronosphere that lets AI agents query the company's observability platform through the Model Context Protocol. It exposes more than 30 tools spanning logs, Prometheus metrics, traces, events, and configuration entities (dashboards, monitors, SLOs, recording/rollup/drop/mapping rules, and notification policies), giving an agent read access to the same data engineers use during incident response and capacity reviews.

The server can be consumed two ways. The easiest is the remote hosted endpoint at https://<org-name>.chronosphere.io/api/mcp/mcp, which any MCP client can connect to with a Chronosphere API token or OAuth (depending on client support). The repo also ships a Go binary (chronomcp) you can build from source and run locally against a YAML config file. Operators can selectively disable tools per-session via the X-Chrono-MCP-Disable-Tools header.

Notable capabilities include PromQL instant and range queries, PNG chart rendering of range queries (render_prometheus_range_query), log histograms, field/label discovery for both logs and metrics, and monitor status tracking. The project is Apache-2.0 licensed and currently pre-1.0, so APIs may shift.

Tools

Prerequisites

• A Chronosphere account with API access
• A Chronosphere API token (or an MCP client that supports OAuth)
• Your organization name (the <org-name> portion of your Chronosphere URL)

Option 1: Remote server (recommended)

Point your MCP client at the hosted endpoint. For Cursor or VS Code, add to your MCP config:

{
  "mcpServers": {
    "chronosphere": {
      "url": "https://<org-name>.chronosphere.io/api/mcp/mcp",
      "headers": {
        "Authorization": "Bearer <api-token>",
        "X-Chrono-MCP-Disable-Tools": "<optional-comma-separated-list>"
      }
    }
  }
}

For Claude Code:

claude mcp add -t http \
  -H "Authorization: Bearer ${CHRONOSPHERE_API_TOKEN}" \
  chronosphere "https://${CHRONOSPHERE_ORG_NAME}.chronosphere.io/api/mcp/mcp"

Option 2: Build and run locally

Clone the repo and build the binary:

make chronomcp

Then configure your client to launch it:

{
  "mcpServers": {
    "chronosphere-mcp": {
      "command": "<PATH/TO/REPO>/bin/chronomcp",
      "args": ["-c", "<PATH/TO/REPO>/config.yaml"],
      "env": {
        "CHRONOSPHERE_ORG_NAME": "<your-org>",
        "CHRONOSPHERE_API_TOKEN": "<your-token>"
      }
    }
  }
}

Environment variables

• CHRONOSPHERE_ORG_NAME: your Chronosphere organization name
• CHRONOSPHERE_API_TOKEN: API token used for Bearer authentication

• Incident triage from chat: ask an agent to pull the last hour of error logs for a service, render a PromQL latency chart, and correlate with recent events.
• PromQL exploration: have an LLM iterate on a query by inspecting label names/values and series metadata before running query_prometheus_range.
• Monitor and SLO review: list active monitors, get their status, and pull the SLOs and recording rules they depend on for a weekly reliability report.
• Log field discovery: when onboarding to a new service, use list_log_field_names and list_log_field_values to map what is actually indexed.
• Config audits: enumerate dashboards, drop rules, mapping rules, and notification policies to answer questions like "which drop rules touch metric X?"

• "Run a PromQL range query for rate(http_requests_total{service=\"checkout\"}[5m]) over the last 6 hours and render it as a chart."
• "Show me the top error log fields for the payments service in the last 30 minutes."
• "List all monitors that are currently firing and pull the recording rules they reference."
• "Find traces tagged env=prod with latency over 2 seconds in the past hour."
• "What events were recorded between 14:00 and 15:00 UTC, and which dashboards reference the affected service?"

• Official server maintained by Chronosphere with broad coverage (30+ tools) across logs, metrics, traces, events, and config.
• Available as a hosted remote endpoint, so most users can skip building from source.
• Supports both API token and OAuth authentication, plus per-session tool disabling via header.
• Apache-2.0 licensed and open source, making it easy to audit or fork.

• Read-only: there are no tools to create, edit, or delete monitors, dashboards, SLOs, or rules.
• Pre-1.0 release, so tool names and behavior may change.
• Requires a Chronosphere account and API access, which is a paid commercial product.

• Grafana MCP server for Grafana, Loki, Tempo, and Mimir users.
• Honeycomb MCP server for event-based observability on Honeycomb.
• Prometheus MCP server for direct PromQL access against any Prometheus-compatible endpoint.