OpenSEO in Claude: what its MCP server does and what Linkeddit adds

OpenSEO describes itself on GitHub as an open source alternative to Semrush and Ahrefs, under the MIT license. Its hosted MCP server lets an AI client call OpenSEO tools for keyword research, SERP inspection, local business research, competitor research, domain research, backlink overview, saved keywords, rank tracking, shared project context and Google Search Console performance and URL inspection.

The Agent Index listing shows 46 tools and auth required. They include list_projects, research_keywords, get_keyword_metrics, find_serp_competitors, get_ranked_keywords, get_backlinks_overview, get_rank_tracker, get_search_console_performance, inspect_urls, run_site_audit and get_local_rank_grid. The first connection sends you through an OpenSEO login, and the OpenSEO docs add that an API key works for headless setups.

OpenSEO also publishes agent skills for Claude Code and Cowork, such as keyword research, keyword clustering, competitor analysis, link prospecting and an SEO report. Those are OpenSEO's own workflows. This page covers the data side and what to pair it with.

Claude can research keywords with volume, difficulty and CPC, fetch live Google results, find the keywords a domain ranks for, compare SERP competitors across a keyword set and read rank tracker positions. OpenSEO's docs list each of these as MCP capabilities. Ask Claude to call list_projects first, because OpenSEO ties a connection to the workspace that was active when you authorized it.

For competitor work, find_serp_competitors compares who ranks across keywords you supply, and get_ranked_keywords and get_backlinks_overview describe a rival's footprint. OpenSEO needs your list to start. Linkeddit can supply it: on its test project, plausible.io, search_competitors_get returned 358 competing domains from 59 checked keywords, with each domain's top 10 keywords and AI answer citations.

Rank tracking lives in OpenSEO's tracker, which Claude reads with get_rank_tracker. Linkeddit's Search tool keeps its own weekly Google ranks and records the AI Overview beside each. On the test project it tracks 25 keywords. For google analytics alternative the project ranked 2, an AI Overview was shown and it cited the project. Run both if you want a second reading.

Both products record AI answers, so choose on what you need beside them. OpenSEO's AI Visibility lets you research prompts, try one in ChatGPT, Claude, Gemini or Perplexity, and track saved questions, with ChatGPT and Gemini selected by default and Google AI Overviews, Claude and Perplexity also supported. Its docs say research questions come from DataForSEO, built mostly from Google People also ask, and are not logged ChatGPT prompts.

Linkeddit records where ChatGPT, Gemini, Perplexity, Claude, Google AI Overviews and Google AI Mode recommend a competitor, lists the cited sources, and turns each gap into an action. It also tracks what competitors changed on blogs, changelogs and pricing pages, and stores competitor reviews for Claude to read. For a team that wants the SEO data and the competitor intelligence together, pair OpenSEO with Linkeddit at $60 a month.

OpenSEO's docs are candid about the limits: tracked ChatGPT and Gemini answers come from the consumer sites, while Claude and Perplexity are asked through their model APIs at live API pricing. Linkeddit shows an estimate before anything that spends credits, and Claude should wait for your yes.

Start from the cited pages, not the keyword list. Linkeddit records, per engine, how often a domain was cited in the latest finished answer run. On the test project, matomo.org was cited 23 times, 21 by ChatGPT, and umami.is 17 times. Claude can then use OpenSEO's get_backlinks_overview on those domains to see who links to them.

Backlink data comes from OpenSEO. Citation counts come from Linkeddit. Ask Claude to keep them in separate columns. A domain that is both cited often and linked from many review sites is a different target from one that is cited and barely linked, and the pages the engines cite are where your answer should be better.

The gap is easy to describe and hard to see by hand. "Tried maybe 20 different phrasings, they're in almost every answer, we appear in like 2," said a startup founder on Reddit. A recorded count per engine replaces trying phrasings one at a time.

Some OpenSEO tools spend credits or change a project. The Agent Index classifies 6 of its 46 tools as writes and 2 as admin, and the docs say Prompt Research and uncached Explorer answers use credits. Ask Claude to preview cost with estimate_rank_tracker_cost before it starts a tracker, and to read before it saves anything.

API keys act as you in the workspace that was active when you created them, according to OpenSEO's docs, so create them with care and revoke them in Settings. Then read the output. The data is only as good as the keyword set you supplied, so check that Claude used your real list.