The API and MCP server for SEO and GEO.
CitationBench is the REST API and MCP server for SEO and GEO — keyword research through rank tracking, content production through link-building, traditional SEO alongside AI search citation monitoring. Built for teams running multiple brands.
Get started in 60 seconds
One command to connect to the MCP server for SEO and GEO. Demo mode requires no key.
# Step 0 — try without signing up (demo mode, no key required)
claude mcp add citationbench https://mcp.citationbench.com/mcp
# Step 1 — connect with your API key
claude mcp add citationbench https://mcp.citationbench.com/mcp \
--header "Authorization: Bearer sk_live_***" \
--header "X-Workspace-Id: ws_***"
# Step 2 — ask naturally in Claude Code
> Research keywords for acme.com, PURCHASE intent, limit 50.
> Bootstrap full SEO presence for newclient.com — pause for my approval at each step.
> Check which of our top 20 keywords are cited in ChatGPT this week.✓ research.keyword.research → 50 keywords labeled (intent × relevance)
✓ agent.invoke bootstrap_brand → running... awaiting approval at step 3
✓ research.ai_citation.check → 7/20 keywords cited · share-of-voice: 34%Demo mode returns shape-complete synthetic data. Real workspaces require an API key.
Six tools. Zero single source of truth. No answer when clients ask about AI search.
The 2026 SEO + GEO stack is fractured. CitationBench replaces it with one operating layer that knows how every step connects.
The Agency Problem
Onboarding a new client takes 3–5 days across Ahrefs, Surfer, Apollo, and GSC — before a single piece of content is produced. And clients are asking why their brand isn't in ChatGPT with no dashboard to show them.
The Enterprise Problem
Six SEO tools that don't share data. Manual reporting cycles. Leadership wants one number for SEO + GEO ROI; three dashboards return three answers. No unified view of what's ranking vs. what's being cited in AI search.
The Developer Problem
Building an SEO agent means wiring up DataForSEO, Ahrefs, GSC, Apollo, and Instantly separately — months before you write agent logic. The MCP ecosystem for SEO was empty. Roll your own or go without.
The complete ops loop
Keyword research, content, indexing, link-building, and rank tracking — one platform.
Every step in one platform. Every step available as a REST endpoint and an MCP tool.
Research
Keyword research with 2D intent × relevance labels, SERP analysis, competitor backlinks, ICP generation, Reddit pain mining, AI citation tracking, content gap analysis.
Produce
Blog posts, landing pages (1 to 10,000), content refiners, content evaluation, OG image generation. Refiners apply your voice to anything generated.
Publish
WordPress, Wisp, Ghost, custom REST. Auto-indexing fires on every publish — Google Search Console + IndexNow in one step.
Index
GSC URL submission + IndexNow on every publish. No manual step. Track which URLs are indexed and when.
Link Build
SERP-to-outreach (1 keyword → 50 drafts), competitor backlink outreach, autonomous inbound negotiation, full link-CRM with Apollo discovery and Instantly send.
Measure
Daily rank tracking, AI citation share-of-voice across ChatGPT, Claude, Perplexity, and Gemini. Portfolio-wide dashboards. One unified visibility score per brand.
Run 50 client brands with the team you have today.
20-minute brand bootstrap
Hand a new client's domain to the bootstrap_brand agent. In 20 minutes you get the ICP, the keyword universe (800–1,500 labeled keywords), the competitor map, Reddit pain points, a 12-week content plan, and 6 landing-page briefs. Was: 3–5 days of manual research.
Multi-workspace agency architecture
One master API key. Unlimited client workspaces. Every brand isolated: its own keywords, content, campaigns, rank data, approval queue. Switch between clients with a single header — X-Workspace-Id: ws_***.
Client-approval gates
No content publishes and no outreach sends without your sign-off. Every agent action that touches the outside world can be held in an approval queue — reviewed in the dashboard or via API. Fully auditable, idempotent, replayable.
AI search citation tracking across ChatGPT, Claude, and Perplexity. Now you'll know.
AI search is absorbing the queries that used to drive clicks. CitationBench tracks AI citation share-of-voice across ChatGPT, Claude, Perplexity, and Gemini — alongside traditional Google rank tracking — so you have one unified visibility score to show clients.
Track citations across every major AI engine.
Every keyword you care about gets queried daily against ChatGPT, Claude, Perplexity, and Gemini. We report whether your domain is cited, mentioned, or absent — plus a share-of-voice score relative to competitors. Surface trends over 30, 60, and 90-day windows.
- ChatGPT, Claude, Perplexity, Gemini coverage
- Citation, mention, and absence states tracked separately
- Share-of-voice scored against your competitor set
SEO + GEO in one dashboard. Not two tools you have to reconcile.
Profound and Otterly track GEO only. Ahrefs and Semrush track SEO only. CitationBench is the only platform that tracks both in one view — because your clients care about both, and the queries overlap.
- Google position next to AI citation status — per keyword, per brand
- Cross-portfolio dashboards for agencies running many clients
- Daily, automated, replayable
MCP Server
Every tool. One MCP server for SEO and GEO.
The only hosted MCP server purpose-built for SEO and GEO. Connect once from Claude Code, Cursor, or any MCP client. About 35 tools across every pillar — research.keyword, produce.blog_post, indexing.gsc, link_building.serp_outreach, agent.invoke — ready to use in your agent today.
Demo mode runs without an API key, so your agents can be built and integrated before any user signs up. Every tool returns shape-complete synthetic data on demo keys.
What an autonomous SEO agent can do with CitationBench today.
Research at scale
research.keyword.research, research.serp.fetch, research.ai_citation.check, research.discuss.research — every research tool is a named MCP tool. Ask Claude to build a keyword universe and it calls these directly.
Produce and publish
produce.blog_post.create, produce.landing_page.create, produce.publish.send — your agent writes the brief, generates content, and publishes to WordPress. One conversation, end-to-end.
Run full agents
agent.invoke — call a named agent (bootstrap_brand, rank_monitor, link_hunter) from your MCP client. The agent orchestrates everything; you approve the gates.
Get started
From sign-up to first result in five minutes.
For agencies and teams
1. Create your account
Sign up, get your master API key, add your first client workspace via POST /v1/workspaces.
Get API key2. Run bootstrap_brand
Hand the client's domain to the agent. In 20 minutes you have ICP, keywords, competitor map, content plan, and landing-page briefs.
Read the playbook3. Approve and publish
Review the agent's work in your approval queue. Approve in bulk. Auto-indexing fires on every publish.
Approval workflowsPlaybooks
Built-in playbooks for the workflows that matter most.
Step-by-step recipes that compose tools across pillars. Every playbook is callable as a sequence of API calls, runnable as an agent, or usable from Claude Code.
Bootstrap a brand's SEO presence
Hand the agent a URL. ICP, 1,000+ labeled keywords, competitor map, Reddit pain points, content plan, and 6 landing-page briefs in 20 minutes.
Read playbookWeekly rank checks across all clients
One scheduled call against every workspace in your account. Email or webhook deltas for rank drops worth acting on.
Read playbookTrack ChatGPT & Perplexity citations daily
Auto-query AI engines against your tracked keywords. Surface share-of-voice deltas, new mentions, and competitor citation drops.
Read playbookTurn a SERP into 50 outreach emails
Point the agent at a keyword. It scrapes the SERP, finds the linking domains, discovers contacts, drafts personalized outreach. Approve in bulk.
Read playbookGenerate 100 landing pages overnight
Programmatic SEO at scale. Feed a keyword list, the agent generates and publishes pages with auto-indexing baked in.
Read playbookBuild an SEO agent in Claude Code
Connect once, then write your own agent loop using research, produce, and link-building MCP tools. Full developer recipe.
Read playbookPricing scales with the size of your portfolio.
We price per workspace and per call volume — and we tune the quote to your team's shape. Tell us how many brands you run and we'll come back with numbers.
The SEO ops platform your team has been stitching together from six tools. Now it's one.
No lock-in. 12-month deprecation window on API changes. SOC 2 in progress. Built on 18 months of compounding SEO + content infrastructure.