getdr

getdr-mcp

getdr-mcp is a small open-source MCP server by Index Pages that gives AI agents one tool, get_dr: Ahrefs Domain Rating (DR) for domains or URLs. It is a thin front for getdr – it asks a getdr server or the getdr binary and makes no Ahrefs requests of its own. Stdio transport, no runtime dependencies. Docs: index-pages.com/getdr.

Setup

Add it to your MCP client (Claude Desktop, Claude Code, Cursor and others use this shape):

{

"mcpServers": { "getdr": { "command": "npx", "args": ["-y", "getdr-mcp"] } } }

That uses the getdr binary on your PATH. To use a getdr server on your network instead:

{

"mcpServers": { "getdr": { "command": "npx", "args": ["-y", "getdr-mcp"], "env": { "GETDR_API": "http://lan-host:8080", "GETDR_SERVE_TOKEN": "s3cret" } } } }

Requires Node.js 18 or later.

Backends

Checked on every call, in this order:

1. GETDR_API is set – one POST $GETDR_API/dr to that getdr server (getdr serve), with Authorization: Bearer $GETDR_SERVE_TOKEN when the token is set. The server holds the Ahrefs key and the shared cache, so this machine needs neither.
2. getdr (0.2.0 or newer) is on PATH – runs getdr --format json -q [--version V] [--fresh] -f - with the targets on stdin, no shell (as arguments after -- when a target starts with #, which -f would skip as a comment). The binary uses its own config, cache and key (AHREFS_KEY), or its own api setting.
3. Neither – the tool returns an error with setup instructions.

Install the binary with go install github.com/index-pages/getdr@latest or from Releases, then get a free Ahrefs API key: index-pages.com/getdr#key.

Both backends return the same records. Calls time out after 120 seconds.

Proxies: the HTTP backend uses Node's built-in fetch, which ignores HTTP_PROXY/HTTPS_PROXY unless NODE_USE_ENV_PROXY=1 is set (Node 22.21 or later, or 24 and later; older versions have no built-in proxy support). The binary backend honours HTTP_PROXY, HTTPS_PROXY and NO_PROXY as usual.

Tool

get_dr – "Ahrefs Domain Rating (0–100) for domains or URLs, via getdr by Index Pages".

Targets are normalized by getdr: a URL becomes its host, www. is dropped, everything is lower-cased and duplicates are removed. Results follow the input order.

Output

The text result is a JSON array with one record per target, and nothing else:

[

{"target":"ahrefs.com","dr":91,"version":"dr3","checked_at":"2026-10-28T12:00:00Z","cached":true}, {"target":"nodata.example","dr":null,"version":"dr3","checked_at":"2026-10-28T12:00:00Z","cached":false} ]

The record is the getdr JSON contract, the same as getdr --format json and getdr serve. dr is null when Ahrefs has no data; checked_at is null too when there was no response or the target is invalid. An invalid target (not a domain or URL, or refused by Ahrefs, such as localhost) also has "error":"invalid target"; no other record has an error key. Clients on MCP protocol 2025-06-18 or later also get structuredContent – {"results": […]} – and an outputSchema for it.

Invalid targets and targets without a response are part of a normal result. Failures come back as a tool error (isError: true) with getdr's message:

• no target got a response at all;
• getdr server answered 401 – the token is missing or wrong, set GETDR_SERVE_TOKEN;
• getdr server answered 503, or the binary exited with code 3 – the Ahrefs key is missing or rejected;
• the server is unreachable, the binary failed, or a call timed out.

Supported MCP protocol versions: 2026-07-28 (stateless: server/discover and per-request _meta, no handshake) and the initialize-based 2025-11-25, 2025-06-18, 2025-03-26 and 2024-11-05, served side by side.

Attribution

Domain Rating data comes from Ahrefs and is subject to the Domain Rating License, which requires the attribution "Domain Rating by Ahrefs" wherever you publish the numbers. The tool result carries only the data, with no attribution in it.

getdr is not affiliated with Ahrefs.

No telemetry

getdr-mcp connects to exactly one place: the getdr server in GETDR_API, when you set it. Otherwise it only starts the getdr binary. No telemetry, no update checks, no analytics, no pings. Requests carry the User-Agent getdr-mcp/ (+https://index-pages.com/getdr). You can verify this in the source: the only outgoing request and the only child process are in src/backends.ts.

Development

npm ci

npm run build npm test

The backend tests build getdr and its mock-backed server from the getdr source tree (../bin, or GETDR_SRC) and need Go; without it they are skipped. The live check runs only when AHREFS_KEY is set.

Releasing

Bump the version in package.json, server.json (twice) and src/version.ts, update CHANGELOG.md, then push a tag vX.Y.Z. .github/workflows/release.yml builds, runs the tests against the getdr source at the same tag, checks that every version matches the tag and that the packed tarball holds dist/index.js with its shebang, and publishes with npm trusted publishing (OIDC, no token, provenance included).

Trusted publishing can only be configured for a package that already exists on npm, so the first release is published by hand once: npm login, then npm run release. Right after that, add the trusted publisher on npmjs.com (package Settings → Trusted publishing: GitHub Actions, index-pages / getdr-mcp, workflow release.yml); the workflow has to complete its first publish within 2 days or the configuration expires.

For any manual publish use npm run release (build, tests, check-release, npm publish), never a bare npm publish: with install scripts disabled (ignore-scripts) prepack does not run and the package would ship without dist/. prepack stays as a fallback for setups that run scripts.

The same workflow then publishes server.json to the MCP Registry as io.github.index-pages/getdr-mcp: it waits until npm shows the new version (the registry checks mcpName in the npm package), proves the index-pages namespace with the workflow's OIDC token (mcp-publisher login github-oidc, no secret) and runs mcp-publisher publish with a pinned, checksummed mcp-publisher. A version already in the registry is skipped.

License

MIT, see LICENSE. Made by Index Pages.