What the agent does
The agent establishes the target domain, reporting interval and URL scope before querying Ahrefs. A root domain, a subdomain and one page describe different link profiles, so the report states which one it covers. There is no repository to clone. Before submitting the minimal body below, add the id of your connected Ahrefs MCP server to the actual call's connections array. Rush's catalog names https://api.ahrefs.com/mcp/mcp and notes that queries consume your Ahrefs API units; the prompt alone grants no access to that account.
Through the connection's available read tools, it requests referring domains, backlink records, gained and lost links, and anchor text distribution. It records the requested date range, retrieved fields, pagination and filters so a reader can reproduce the selection. Referring domains and backlinks receive different tables: several links from one site should not look like several independent referring sites. If historical data, domain rating or a needed export is unavailable through the connection, the corresponding section says so rather than substituting a current snapshot for the missing period.
The prompt asks for the twenty strongest referring domains. The agent states how it ranks them using fields actually returned, keeping domain rating separate from topical relevance and the value of the linked page. It compares gained and lost records within the requested 30-day interval and preserves source URL, destination URL, anchor and observation dates where available. A link Ahrefs marks lost is an observation to investigate, not proof that a publisher deliberately removed it. Crawl timing, redirects and page changes remain possible explanations.
For anchor review, the agent separates branded, URL-like and repeated keyword anchors, explains the grouping and links its concerns to the underlying records. An unusual share becomes a review candidate rather than a claim of a search penalty. Reclamation suggestions prioritize useful destination pages and relevant sources, distinguish broken destinations from removed source links, and state what to check next. It writes backlinks.md without editing a site, changing a redirect, contacting publishers or starting an outreach campaign.
Start it with one API call
POST https://api.prix.dev/o/{org}/sessions with this body:
{
"agent": "claude",
"input": "Use the attached Ahrefs MCP connection to review the domain and URL scope I identify in this task, for the last 30 days. If the target domain is missing, report that missing input rather than choosing one. Write backlinks.md with the 20 strongest referring domains under an explicit ranking method, new and lost backlinks with source/destination URLs and domain rating where available, anchor distribution and evidence-backed over-optimization candidates, plus lost links worth investigating for reclamation. Record scope, dates, filters, pagination, API-unit limits and missing fields. Read only; do not edit websites, change redirects or contact publishers."
}- Get an API key. In the console, open Settings, then API keys, and create a key for your org. Send it in the Authorization header as a Bearer token.
- Send the request. POST the body in request.json to https://api.prix.dev/o/{org}/sessions, with your repository as the resources URL. The run starts in its own cloud sandbox.
- Read what came back. Follow the run on its live stream or wait for the run.completed webhook, then read the result and its trajectory.
What comes back
backlinks.md contains a scope note, referring-domain ranking, gained and lost link tables, anchor analysis and a prioritized reclamation list. Each suggested action names the link evidence and the verification it still needs. The report distinguishes measured fields from judgment, and incomplete coverage from an observed absence of links. A missing historical endpoint cannot silently become a claim that no links were lost.
A reviewer checks the domain scope, ranking criteria, duplicate-domain handling and whether important destination URLs changed during the period. They inspect a lost link before assigning outreach or redirect work. The report returns as a session artifact, while the transcript preserves the read queries and outputs used to assemble it. This task requests an analysis file; it makes no promise of a particular ranking improvement or a completed outreach result.
Run it on every monthly link review
Your own scheduler can submit one session per authorized domain with explicit start and end dates in input. Attach the Ahrefs connection and supply the exact target instead of expecting the agent to infer a site from an account name. Use an Idempotency-Key to make retries of that same monthly request safe.
Read the finished session or receive its completion webhook, then retain the report with its filters and scope. Compare it with the prior month's report before treating a ranking change as a new link opportunity. A scheduled review should account for the Ahrefs units its queries consume, especially when covering multiple domains.
The use case this job follows, with its automation and trigger: Produce a report or file, no repo.
Why run it on Rush?
A monthly backlink review needs joins and judgment across domains, individual links, anchors and destination pages. Rush gives the agent a sandbox to reconcile those datasets and write a consistent report without leaving exports scattered across a marketer's laptop. The session can return a useful file even though no source repository is involved.
The captured queries make the report's scope checkable when a reader asks why a domain appears or why a lost link ranks highly. Separate domain reviews can run within your API key's concurrency allowance, each with its own input and result. That keeps different sites' scopes and source evidence apart while your server collects the completed artifacts.
Frequently asked questions
Does the report need a GitHub connection?
No, the work reads Ahrefs data and writes a report, so resources is omitted. You do need to attach the Ahrefs MCP connection through connections on the submitted API body. The minimal prompt shown here does not attach it automatically.
Will it consume Ahrefs API units?
Yes, the connector catalog states that Ahrefs queries use your API units. Ask for the scope and fields you need, and have the agent report when access or unit limits interrupt collection. The report should not imply full coverage after a partial retrieval.
Does a repeated anchor prove over-optimization?
No, repeated branded or navigational anchors may be expected. The agent groups anchors, shows the relevant distribution and identifies records worth reviewing. It does not diagnose a search penalty from an anchor percentage.
Will the agent ask a publisher to restore a link?
This request stops at a reclamation recommendation. It can explain which source and destination to inspect and why the link matters. Contacting a publisher or changing a website requires a separate authorized action.