Datadog MCP Server

Community

Read-only access to Datadog monitoring data via MCP

How to install ↓ Compare GitHub source repository ↗

Category: Dev Tools #95 of 438
Stars: ★ 69 Popular
Transport: stdio (local process)
Runtime: Node.js 16+
Credentials: API key / credential required
License: MIT
Last commit: 9 months ago ⚠
Tools: 10

62FMRS · C

A community-maintained MCP server, documented in the README, offering read-only access to monitors, dashboards, metrics, events, logs, and incidents via Datadog's official APIs. The tool set is clear, regional endpoints and scoped keys are supported, and security guidance is sensible — but it is a third-party project with no write capabilities.

Strongest · Documentation 16/20 Weakest · Reliability 7/20

Reliability

7/20

Security and permissions

15/20

Maintenance

10/20

Documentation

16/20

Setup experience

14/20

Why each score

Reliability 7/20: The README lists 10 read-only tools (get-monitors, get-dashboards, get-metrics, get-events, get-incidents, search-logs, aggregate-logs, etc.) and claims 'comprehensive error handling,' but no test files, CI workflow, or execution logs are provided as evidence, and the manifest was not cached, so declared tool behavior cannot be verified against real behavior. Per the static-review rule, reliability cannot exceed 12 without verifiable test/CI evidence; scored 7 to reflect a plausible-but-unverified happy path.
Security and permissions 15/20: All documented tools are read-only (no delete/write/deploy actions), and the README explicitly maps each tool to its minimum required Datadog scope (e.g., monitors_read, logs_read_data) and recommends scoped Application Keys — solid least-privilege guidance. Example credentials are placeholders only, with no real secrets. Deductions: API/App keys can be passed as CLI arguments (--apiKey=...), which risks exposure via process listings on shared systems, and this is not flagged; data-flow disclosure about what happens to log/metric data after it leaves Datadog (caching, retention) is also absent. Scored 15.
Maintenance 10/20: MIT licensed, not archived, with 69 stars and 7 open issues indicating some engagement, but the README gives no information on release cadence, recent commit activity, dependency-update practice, or a security-response channel, so activity and governance cannot be confirmed. Scored 10 per the rule that missing evidence is a deduction, not an assumption of best case.
Documentation 16/20: Documentation is fairly layered: prerequisites, an Application Key scope table, npm and from-source installation, env-var and CLI configuration, regional endpoint list, Claude Desktop and MCP Inspector integration examples, a full tool list, 4 worked examples, and a troubleshooting/debugging section. Gaps: only 4 of 10 tools have concrete parameter examples, and there is no full parameter schema, rate-limit, or cost disclosure, so full marks are not warranted. Scored 16.
Setup experience 14/20: The setup path is clear: prerequisites (Node 16+, Datadog API/App keys), npm global install or from-source build, two configuration options (env vars or CLI args), regional site notes, and a complete Claude Desktop JSON config example. Since no CI or test execution evidence is present in a static review, setup is capped at 15; combined with a clear but manual (credential-entry-dependent) flow lacking automated verification, scored 14.

Static review · not runListed 2026-09-07

Read the FMRS scoring method →

Fit and risk

What it can accessUses the network

Best for

Ops and SRE teams using Claude Desktop who need read-only access to Datadog data
Developers who want monitors, logs, and metrics queries inside LLM workflows
Teams practicing least privilege who can scope their Application Keys

Not for

Scenarios requiring write operations to Datadog (creating/modifying monitors, muting, etc.) — the server offers read-only tools only
Users without a Datadog account, API key, and Application key
Those requiring official Datadog support — this is a community project, not official

Required permissions

DD_API_KEY: Datadog API key
DD_APP_KEY: Datadog Application key, recommended scopes per feature: monitors_read, dashboards_read, metrics_read, events_read, logs_read_data, incident_read
Application Keys without explicit scopes inherit all permissions of the creating user

Risks and side effects

Leaked API or Application keys expose organizational monitoring and log data
Application Keys without scopes carry the creator's full permissions, violating least privilege
Third-party community project, not audited by Datadog
Logs and metrics may contain sensitive business data entering LLM context — consider compliance implications

Setup

Before you start

Runtime：Node.js 16+

DD_API_KEY requiredsecret Datadog API key, found in Organization Settings > API Keys.

DD_APP_KEY requiredsecret Datadog Application key, found in Organization Settings > Application Keys; scopes are recommended for least privilege.

Other optional settings (3)

DD_SITE optional Datadog site domain (e.g. datadoghq.com, datadoghq.eu, us5.datadoghq.com); optional, defaults to datadoghq.com.

DD_LOGS_SITE optional Optional dedicated site for the logs endpoint; defaults to DD_SITE.

DD_METRICS_SITE optional Optional dedicated site for the metrics endpoint; defaults to DD_SITE.

Prepare Node.js 16+ and a Datadog API key and Application key (created in Organization Settings).
Install globally via npm: npm install -g datadog-mcp-server, or build from source with npm install && npm run build.
Add an mcpServers entry to Claude Desktop's claude_desktop_config. using npx datadog-mcp-server with --apiKey, --appKey, and --site arguments; the config file is at ~/Library/Application Support/Claude/ on macOS or %APPDATA%/Claude/ on Windows.
Alternatively debug with MCP Inspector: npx @modelcontextprotocol/inspector datadog-mcp-server --apiKey=... --appKey=...

claude_desktop_config.json

{"mcpServers":{"datadog":{"command":"npx","args":["datadog-mcp-server","--apiKey","<YOUR_API_KEY>","--appKey","<YOUR_APP_KEY>","--site","<YOUR_DD_SITE>"]}}}

Shown for Claude Desktop. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.

.vscode/mcp.json

{
  "servers": {
    "datadog": {
      "command": "npx",
      "args": [
        "datadog-mcp-server",
        "--apiKey",
        "<YOUR_API_KEY>",
        "--appKey",
        "<YOUR_APP_KEY>",
        "--site",
        "<YOUR_DD_SITE>"
      ]
    }
  }
}

Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).

Terminal

claude mcp add datadog -- npx datadog-mcp-server --apiKey '<YOUR_API_KEY>' --appKey '<YOUR_APP_KEY>' --site '<YOUR_DD_SITE>'

Run it in a terminal; replace any <…> placeholders with your own values first.

Generate and validate client config

Check that it works

After connecting in Claude Desktop, confirm tools like get-monitors and search-logs appear in the client's tool list; then ask to list your Datadog monitors — returned monitor data proves the connection works.

Troubleshooting

403 Forbidden: verify API and Application keys are correct, have required permissions, and the account can access the data
Confirm the correct regional endpoint (e.g., datadoghq.eu for EU customers)
Invalid key format errors: ensure you are using the full key strings
Endpoint mismatches: if logs and metrics use different domains, check --logsSite and --metricsSite settings
Check Claude Desktop MCP logs (macOS: ~/Library/Logs/Claude/mcp*.log, Windows: %APPDATA%/Claude/Logs/)

Things to try

Once connected, you can ask your AI assistant things like:

List my Datadog monitors, showing the first 5 in alert or warn state
Show me the dashboard definition for dashboard ID abc-def-123
Search logs for service:web-app errors from the last 15 minutes
Count logs for service:web-app over the past hour, grouped by status

Tools 10

get-monitors read-only

Fetch monitors with optional filtering

get-monitor read-only

Get details of a specific monitor by ID

get-dashboards read-only

List all dashboards

get-dashboard read-only

Get a specific dashboard by ID

get-metrics read-only

List available metrics

get-metric-metadata read-only

Get metadata for a specific metric

get-events read-only

Fetch events within a time range

get-incidents read-only

List incidents with optional filtering

Show 2 more tools

search-logs read-only

Search logs with advanced query filtering

aggregate-logs read-only

Perform analytics and aggregations on log data

Use cases

Query Datadog monitor states and configurations in conversation

Inspect dashboard definitions by ID

Search and aggregate logs for troubleshooting

Look up metric metadata and event stream data

Review active incidents

Supported clients

Claude Desktop

Listed from the project's documentation, not tested by this site.

Overview

Datadog MCP Server is a Model Context Protocol server that provides read-only access to monitors, dashboards, metrics, events, logs, and incident data through Datadog's official v1 and v2 APIs. It supports multiple Datadog regional endpoints (US, EU, US3, US5, AP1) and separate endpoints for logs and metrics. Configuration is via environment variables (DD_API_KEY, DD_APP_KEY, DD_SITE, etc.) or command-line arguments. This is a community-maintained project (GeLi2001/datadog-mcp-server), not an official Datadog product, licensed under MIT.

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 63.3k · Tools 2 Compare with this →

NPM Sentinel MCP Server 79 · B

AI-powered NPM package analysis MCP server

★ 18 · Tools 19 Compare with this →

MartinLoop MCP Server 79 · B

An independent control layer that gives AI coding agents budgets, brakes, and receipts

★ 413 Compare with this →

GitHub MCP Server 76 · B

GitHub's official MCP server for managing repos, issues, PRs, and workflows via natural language

★ 33.7k · Tools 15 Compare with this →

Compare with other MCP servers

Source revision 8665aa40deb7 Data synced 2026-10-19 Read the FMRS scoring method