OpenSEO MCP server
Commit Oct 31, 2026Lets an AI client perform SEO audits, keyword research, rank tracking, and backlink analysis using 76 specialized tools.
Also known as duckduckdeploy/openseo
- Stars
- 50
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- —
- Last commit
- Oct 31, 2026
- License
- MIT
What is OpenSEO MCP?
OpenSEO lets an AI client perform SEO audits, keyword research, rank tracking, and backlink analysis using 76 specialized tools.
Summary generated from the project's README
About OpenSEO MCP server
Performing SEO audits, keyword research, and rank tracking from inside an AI client is what OpenSEO enables. This open-source suite provides 76 specialized MCP tools that allow agents to analyze domains, track keyword positions, and examine backlink profiles. It functions as an alternative to traditional paid SEO platforms by offering a model where the software is free and search data is paid per lookup.
Users can run the entire suite on their own machine or server using Docker, ensuring that all keywords, rankings, and audits remain in a private database. To prevent unexpected costs, every paid tool includes parameters like dry_run and max_cost_usd so an agent cannot exceed a set budget. The system also integrates with Google Search Console and Analytics to provide insights into search performance and opportunity scores. Beyond standard web search, it includes specialized capabilities for tracking AI visibility across chat models and local rank tracking via geo-grid maps.
What you can do with it
- 1Conducting domain SEO auditsAn AI agent can perform a full audit of any domain to identify important findings, traffic, and authority.
- 2Performing keyword researchDevelopers can use the server to generate keyword ideas including volume, difficulty, and intent.
- 3Monitoring AI visibilityUsers can track how often their domain is mentioned or cited across various chat models.
- 4Tracking local search rankingsAn agent can generate geo-grid rank maps to see how a business appears on Google Maps.
Written from the project's README. Every use case is checked against a quote from it.
Automated checks
Ran Nov 15, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachableduckduckdeploy/openseo is public and reachablegithub.com/duckduckdeploy/openseo ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)MITspdx.org/licenses/MIT.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-10-31github.com/duckduckdeploy/openseo ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHubOpenSEO (openseo.gg) is an open-source alternative to Ahrefs and Semrush. It covers full SEO audits of any domain, keyword research, rank tracking, backlinks, technical site audits, AI visibility and local rank tracking, and reads your own Google Search Console and Analytics data. Run it on your own machine or server, read every line of it, and hand it to your AI agent.
No monthly plan: the software is free, and search data is paid per lookup, under a monthly limit you set. A rank check is from $0.02, a keyword search of 150 ideas about $0.08.
⭐ If OpenSEO is useful to you, a star helps other people find it.
[!NOTE] OpenSEO is released and gets active updates. Issues and ideas are very welcome.
📸 A look inside
Every tool is one search: type a keyword, domain or URL, pick a country (and a city or state where it helps), press Enter, and everything loads together. The search is in the address, so any view can be bookmarked or shared. Each tool has its own page, and the guide goes through every screen.
🆓 Try it free, no account
Six tools run right on the website, a few free checks a day, no sign-up:
| Tool | What it checks |
|---|---|
| Keyword generator | 50 keyword ideas for a seed, with volume, difficulty and CPC |
| Keyword difficulty checker | Difficulty of up to 20 keywords |
| SERP checker | The live Google top 20 for a keyword |
| Website traffic checker | Estimated Google traffic of up to 5 sites |
| Website authority checker | Domain Rating, referring domains and backlinks of up to 3 sites |
| AI Overview checker | Whether Google shows an AI Overview, and which sites it cites |
✨ Why OpenSEO
- 🔓 Open source, all of it. The app, the API, the crawler and the agent server are MIT licensed. Change what you need and keep your changes.
- 🏠 Your server, your data. One Docker command runs the whole suite on your own computer or server. Your keywords, rankings and audits stay in your own database.
- 💸 No surprise bills. Every call is listed with its cost, and a monthly limit stops all spending once it is reached.
- 🤖 Agents on a leash. 76 MCP tools cover the whole workspace. Every paid tool takes
dry_runandmax_cost_usd, so an agent cannot spend more than you agreed. - 📈 Your own Google data. Connect Search Console. You get queries one push from page one, and pages scored by where improving them is most worth it.
- 🎯 Focused workflows. The SEO work most sites actually do, without the hundred reports nobody opens.
🧰 Workflows
🥊 OpenSEO vs the paid suites
| Ahrefs, Semrush, SE Ranking, Moz | OpenSEO | |
|---|---|---|
| Source code | Closed | Open source (MIT) |
| Where it runs | Their cloud | openseo.gg, or your own server with Docker |
| Price | Monthly plans from $49 (Moz) to $140 (Semrush) to start | Free software; data paid per lookup |
| Seats | Paid per user | No seats |
| AI agents | API or MCP on some plans | 76 MCP tools and a REST API, spending capped per call |
| Data depth | Their own crawlers and indexes, the deepest | A large commercial dataset through a provider |
Honest side-by-sides, including when each one is the better buy: Ahrefs · Semrush · SE Ranking · Mangools · Ubersuggest · Moz · Screaming Frog · Otterly.AI
🐳 Self-host with Docker
You need Docker. On Windows or macOS that's Docker Desktop with at least 4 GB of memory.
git clone https://github.com/duckduckdeploy/openseo.git openseo
cd openseo
cp .env.example .env
# In .env, set AUTH_SECRET (32+ random characters)
# and ENCRYPTION_KEY (openssl rand -base64 32)
docker compose up -d --build
| Service | Address |
|---|---|
| Dashboard | http://localhost:3200 |
| Site and docs | http://localhost:3101 |
| API and MCP | http://localhost:4100 (/api/*, /mcp) |
One image runs everything: Postgres, the API with its MCP server and background worker, the dashboard and the site. Database migrations run when the API starts.
On a cloud server with your own domain, add HTTPS with automatic Let's Encrypt certificates:
- Point your domain,
app.andapi.at the server. - Set
OPENSEO_DOMAIN,APP_URL,WEB_URL,PUBLIC_API_URLandPOSTGRES_PASSWORDin.env. - Run
docker compose --profile https up -d --build.
The self-hosting guide covers servers, updates, backups and every setting.
🤖 Connect your AI agent
The SEO MCP server gives agents the whole workspace as 76 tools, listed with their prices in MCP tools.
ChatGPT and Claude (web and desktop): add a custom connector with the URL {API_URL}/mcp and OAuth sign-in. You sign in with your OpenSEO account and approve; no key needed. Connected apps can be revoked in Settings.
For other agents, create an API key under Settings and store it in an environment variable named OPENSEO_API_KEY.
The quick way: copy the setup prompt from Settings and paste it into your agent. The agent connects itself and checks the connection with free calls only.
Claude Code: install the plugin, which bundles the connection and the skills:
/plugin install openseo --marketplace duckduckdeploy/openseo
Set OPENSEO_MCP_URL to point it at a self-hosted install.
Any other MCP client: add {API_URL}/mcp with the header Authorization: Bearer <key>. The MCP guide has the exact config for:
- Claude Code and Claude Desktop
- Cursor
- VS Code
- Codex
- Gemini CLI
- OpenCode
- Windsurf
🧩 Skills
Ready-made SEO workflows in plugins/openseo/skills, built on the 76 MCP tools:
| Skill | What it does |
|---|---|
seo-coach |
Learn SEO on your own site, and pick the first thing worth doing |
seo-audit |
Any domain's rankings, links and competitors, as a short plan |
site-audit |
Crawl your site: the one or two technical fixes that matter most |
keyword-research |
Topics worth writing about, grouped and saved |
competitor-analysis |
Their strategy, the keyword gap, and where you can win |
search-console-review |
Quick wins from your own Google data, free |
local-seo |
Your Maps profile, reviews and rankings street by street |
Every skill prices paid steps with dry_run first and passes the limit you agree as max_cost_usd. In Claude Code, /seo-audit example.com audits any domain in one call (the seo_audit tool, about $0.50 to $0.65) and turns it into a short plan. Or just ask, for example "which of my pages are stuck on page two?". The same features are available over the REST API.
🔎 Search data
The crawler, site audit and Google data are free. Keyword volumes, rankings, backlinks and other search data come from YepAPI. You add your own key when you set up your account, and each lookup is billed to it at the price the app shows first: a page-one rank check costs $0.01, and 150 keyword ideas cost about $0.08.
🛠️ Development
Requirements: Node 26+, pnpm 10, Postgres 15+ (or docker compose up -d postgres).
pnpm install
pnpm db:migrate
pnpm dev # API :4100, dashboard :3200, site :3101, with live reload
pnpm typecheck
pnpm test
The tests need no database, network or API keys. apps/api/test runs the real API against an in-memory Postgres (PGlite) with the real migrations, plus stand-ins for the data provider (apps/api/src/testing/fake-yepapi.ts) and Google (apps/api/src/testing/fake-google.ts). For local work without a data key, run the stand-in with pnpm --filter @openseo/api fake-yepapi and set YEPAPI_BASE_URL=http://localhost:4199.
apps/api Fastify API, MCP server, background worker (pg-boss)
apps/app Dashboard (Next.js)
apps/web Public site and docs (Next.js)
packages/yepapi Typed data client with price estimates
packages/crawler Site-audit crawler
packages/shared Model catalog and shared helpers
User documentation and tutorials live on the site under /docs (apps/web/src/app/docs).
🔐 Security
- Stored data keys and Google tokens are encrypted at rest with AES-256-GCM (
ENCRYPTION_KEY) and never returned by the API. - API keys are stored as SHA-256 hashes and shown once.
- OAuth for MCP clients:
- dynamic client registration and PKCE (S256)
- exact redirect URIs
- single-use codes
- access tokens that last one hour
- rotating refresh tokens; a reused one ends the grant
- tokens stored as hashes
- Paid lookups are not retried after a server error or a lost connection, since they may already have been billed.
- The crawler refuses private, loopback and metadata addresses, including through redirects.
- In Docker, the app, site and API listen only on the machine itself. The optional Caddy service is the only thing facing the internet.
🤝 Contributing
Clear issues are the most useful contribution: what you did, what you expected and what happened. For code, run pnpm typecheck and pnpm test before opening a pull request.
📄 Licence
MIT. Use it, change it, host it.
OpenSEO MCP questions
How much does it cost to use?
The software is free to use, but search data is paid per lookup with a monthly limit you can set.
Can I run OpenSEO on my own hardware?
Yes, you can self-host the entire suite on your own computer or server using Docker.
How can I prevent an AI agent from spending too much money?
Every paid tool includes a dry_run and max_cost_usd parameter so you can cap spending per call.
What kind of SEO tools are available via MCP?
There are 76 MCP tools available that cover the entire workspace.