Atatus MCP Server connects Claude Code, Cursor, Codex, and any MCP-compatible agent to live logs, traces, errors, deployments, and metrics. Debug production issues without switching tools.
Pick a question to see which Atatus tools your assistant calls.
Why is the checkout-service slow this week?
get_service_health_summarydoneanalyze_slow_transactionsdoneget_transaction_spansdonePOST /api/checkout p95 went from 310 ms to 1.2 s. Most of it is one query on order_items, run 38 times per request.
Pick an area, open a tool, then follow its next step the way your assistant chains them.
get_service_health_summaryAPMA HEALTHY / WATCH / DEGRADED verdict plus the golden signals in one call: Apdex, service-wide failed requests and error rate, top open error fingerprints, the top endpoints by request count with their…
Point your client at mcp.atatus.com with an Application Key, then check the connection.
Settings » Account Settings » API Keys. Keep it Read unless you need writes.
claude mcp add --transport http atatus-mcp https://mcp.atatus.com/mcp \ --header "X-API-KEY: ak_xxxxxxxxxxxxxxxxxxxxxxxx"
Sends the X-API-KEY header itself, straight to the hosted server. Nothing to install.
# run a checkListing tools works without a valid key, so a real tool call is the true test.
One tool writes. Switch the key and try it.
update_error_statusLocked. A Read key rejects the write tool. Every other tool still works.
Use the Atatus app for these.
Pricing, deployment, data and the agent. If something is missing, a person will answer it on a demo call.
The Model Context Protocol (MCP) is an open standard that lets AI assistants call tools to fetch data from other systems. The Atatus MCP server is hosted at mcp.atatus.com and gives your assistant 67 tools to query your APM, browser and mobile monitoring, infrastructure, Kubernetes, logs, traces, databases and uptime checks in plain English.
Any client that supports remote MCP over HTTP. The docs have ready-to-paste setups for VS Code, Cursor, Antigravity, Claude Code, Antigravity CLI, Codex CLI and Claude Desktop. Other clients connect by pointing at https://mcp.atatus.com/mcp and sending your key in the X-API-KEY header.
The server is read-only by default. A Read key lets the assistant query everything and change nothing. The only write action is resolving, ignoring or reopening errors, and it needs a Read & Write key. Each key reads only its own account's data, is sent over HTTPS in the X-API-KEY header, and can be revoked at any time under Settings » Account Settings » API Keys.
Yes. Tools such as get_logs, analyze_logs, search_traces, get_trace_detail and get_slowest_spans return logs and traces, and other tools cover errors, deployments, metrics, Kubernetes, hosts and databases. Log queries through the MCP server cover the last 7 days.
Yes. In Claude Code, run claude mcp add --transport http atatus-mcp https://mcp.atatus.com/mcp with your key in the X-API-KEY header. In Cursor, add the server to ~/.cursor/mcp.json for all projects or .cursor/mcp.json for one project.
Under 5 minutes. Create an Application Key under Settings » Account Settings » API Keys, add the Atatus server to your client's configuration, restart the client and ask your first question. The server is hosted, so there is nothing to install. Claude Desktop is the one exception: it needs a small local relay, mcp-remote, which runs through npx and requires Node.js.
Only the data a tool returns to answer your question, and only from your own account. What your AI client does with that data afterwards is governed by that client's privacy and data-handling terms.
Connect Claude Code, Cursor or Codex to live logs, traces, errors and metrics.
Compare Atatus with Datadog