AnswerRange ships as an MCP server: one file, no dependencies. Add it to Claude Desktop, Claude Code or Cursor and ask "measure example.com on these eight questions". You get the share of answers that cite the site, with its interval, per platform and per question, and a link to the public report.
1. Download answerrange-mcp.js (Node 18 or later; nothing else).
2. Register it with your client.
Claude Desktop — claude_desktop_config.json:
{
"mcpServers": {
"answerrange": { "command": "node", "args": ["/path/to/answerrange-mcp.js"] }
}
}
Claude Code:
claude mcp add answerrange -- node /path/to/answerrange-mcp.js
For the drafts tools, add your token: claude mcp add answerrange -e ANSWERRANGE_TOKEN=arp_… -- node /path/to/answerrange-mcp.js, or an "env": { "ANSWERRANGE_TOKEN": "arp_…" } entry next to "args" in the Desktop config.
Cursor — .cursor/mcp.json with the same mcpServers block.
| Tool | What it does |
|---|---|
list_categories | The categories, markets and platforms that can be measured, and the free-run rules. |
start_measurement | Measure a domain on a category's questions, or — better — on 8–12 questions you supply per market. Returns the exact questions that will be asked and the report link. Ten to fifteen minutes. |
get_measurement | Status while running; when done, the headline with its interval, per-platform rates, per-question hits, share of voice, report URL. |
Drafts — need a personal token (drafting plans; make one on your dashboard) set as ANSWERRANGE_TOKEN. | |
list_projects | Your measured domains, and whether you can draft for each. |
get_gap | "What to write next": lost questions and Search Console queries at positions 4–20, each with the question, market and id to draft from. |
create_draft | Draft one page from the store's own facts — internal links, SEO fields, the store's photos, every unknown store fact as a [[CHECK]]. One to two minutes; costs model credit; ten per project per day. |
get_draft | Read a draft (or list them): HTML, FAQ, SEO fields, open checks, notes, audit. |
update_draft | Edit fields, approve (refused while a [[CHECK]] remains), record the URL it went live at, or delete it. Publishing to WordPress stays a button on the draft page. |
The tool descriptions tell the assistant what a zero means: 0 of 108 on generic category questions is a property of the questions, not the site, so it will steer you toward your own questions. It also tells it a run costs money and not to start more than you asked for. Free runs: one per address per hour, two per day; the same guards as the site.
The server is a thin client over three JSON endpoints, which you can call directly.
GET /api/v1/categories
POST /api/v1/measurements
{ "domain": "example.com", "category": "b2b-software", "markets": ["en"],
"questions": { "en": ["best error monitoring for a node app", "..."] },
"email": "you@example.com" }
→ 202 { id, status: "queued", questions: [...], report, poll }
GET /api/v1/measurements/{id}
→ { status, sources, headline: { cited, answers, rate, interval }, platforms: [...], questions: [...], shareOfVoice, report }
# drafts — header Authorization: Bearer arp_…
GET /api/v1/projects
GET /api/v1/projects/{id}/gap → { draftable, rows: [{ group, market, question, qid, … }] }
GET /api/v1/projects/{id}/drafts
POST /api/v1/projects/{id}/drafts { "question": "…", "market": "de", "qid": "site-de-1" }
→ 201 draft | 409 cited pages still being read (retryAfterSeconds) | 429 daily cap
GET /api/v1/projects/{id}/drafts/{draft}
PATCH /api/v1/projects/{id}/drafts/{draft} { "html": "…", "status": "approved", "publishedUrl": "https://…" }
DELETE /api/v1/projects/{id}/drafts/{draft}
Rates are percentages; intervals are 95% cluster bootstrap over questions, exact binomial (Clopper–Pearson) at 0 or n. A platform that could not be measured
(credit exhausted, provider down) is reported as notMeasured with the reason, never as zero.