CitationBench

LLM Visibility MCP Server for Agencies and Developers

CitationBench is a hosted LLM visibility MCP server exposing ~35 SEO and GEO tools. One connection, zero self-hosting, multi-workspace agency support via a single header.

Talk to SalesRead the docs
Claude Code · CitationBench MCP
# Connect CitationBench LLM visibility MCP server> claude mcp add citationbench --url https://mcp.citationbench.com/mcp✓ Connected · 35 tools available# Track AI citations across ChatGPT, Claude, Perplexityresearch.ai_citation.check> { "query": "best seo platform for agencies", "engines": ["chatgpt","claude","perplexity"] }chatgpt    · cited: ✓  rank: 2  share_of_voice: 0.38claude     · cited: ✓  rank: 4  share_of_voice: 0.21perplexity · cited: ✗  rank: —  share_of_voice: 0.00# Switch workspace context with one headerX-Workspace-Id: ws_clientbrand42 · tools ready

LLM Visibility Is Broken for Builders

Fragmented tooling and no-AI-search coverage leave agencies and developers operating blind.

No AI Search Visibility

Traditional SEO platforms track Google ranks but cannot tell you if ChatGPT, Claude, or Perplexity cites your clients.

Months of Integration Debt

Building an MCP-compatible SEO tool layer from raw APIs takes months before a single agent workflow runs in production.

One Server per Client Breaks Scale

Most hosted MCP servers lack multi-workspace architecture, forcing separate setups for every client brand you manage.

The fix

One Hosted MCP Server. Full LLM Visibility.

CitationBench provides a production-ready hosted MCP server at mcp.citationbench.com/mcp exposing ~35 SEO and GEO tools. Connect once from Claude Code, Cursor, or any MCP client. Manage every client workspace via the X-Workspace-Id header. Track AI citations alongside traditional ranks.

Everything Your LLM Agent Needs for SEO Ops

Five production-ready capabilities accessible immediately from any MCP client — no self-hosting, no glue code, no separate tools.

Hosted LLM Visibility MCP Server Ready Now

MCP Infrastructure

Hosted LLM Visibility MCP Server Ready Now

Connect to mcp.citationbench.com/mcp and get ~35 namespaced tools instantly — research.*, produce.*, indexing.*, link_building.*, agent.*. Works with Claude Code, Cursor, Claude Desktop, Windsurf, and ChatGPT Apps.

  • No self-hosting required
  • 35 tools across 5 namespaces
  • Claude Code and Cursor ready
Track AI Citations Across Every LLM Engine

LLM Visibility

Track AI Citations Across Every LLM Engine

Call research.ai_citation.check to monitor brand citations in ChatGPT, Claude, and Perplexity. Returns share-of-voice per engine with historical citation trajectory. Daily query scheduling keeps visibility current.

  • ChatGPT, Claude, Perplexity support
  • Share-of-voice per engine
  • Historical citation trajectory
One API Key for All Client Workspaces

Agency Architecture

One API Key for All Client Workspaces

A single master API key manages N client workspaces. Pass X-Workspace-Id on every MCP tool call to scope keywords, content, campaigns, and link-building to the correct client. White-label-ready.

  • X-Workspace-Id header routing
  • Full per-client data isolation
  • Bulk actions across workspaces
Invoke Full SEO Agents from Your LLM

Agentic Workflows

Invoke Full SEO Agents from Your LLM

POST /v1/agent/invoke runs named agents with structured input directly from your MCP client. Real-time SSE event streaming, multi-turn continuation, and approval-gated pauses keep agents auditable and resumable.

  • SSE event streaming
  • Human approval gate support
  • Multi-turn agent continuation
Durable Jobs That Survive Restarts

Reliability

Durable Jobs That Survive Restarts

Agent jobs run on Cyclonic workers — durable, resumable, and cancellable. Cancel via agent.invocations.cancel, resume via agent.invocations.continue. Not fire-and-forget LLM calls that vanish on timeout or restart.

  • Survives worker restarts
  • Cancellable and resumable
  • Full job audit trail

How it works

From sign-up to first call in five minutes.

01

Sign up and get your key

An `sk_test_*` key lands in your dashboard instantly. No demo gate — start calling real endpoints with shape-complete responses.

02

Add the MCP server or hit REST

`claude mcp add citationbench https://mcp.citationbench.com/mcp` from Claude Code, Cursor, or any MCP client. Or `curl` against `api.citationbench.com/v1/*` directly.

03

Run a tool or invoke an agent

Every tool returns shape-complete demo data without auth, so your agent works before the user signs up. Tools across research, production, indexing, link-building, and agents.

04

Scope to a workspace and ship

Add `X-Workspace-Id: ws_***` to scope per client. Switch to a live `sk_live_*` key when you're ready. Same API surface, same SDK.

Why CitationBench

Built for production, agency-scale, and AI-agent-first.

~35 tools, one MCP server

Hosted at mcp.citationbench.com/mcp. Works with Claude Code, Cursor, Claude Desktop, Windsurf, and ChatGPT Apps. No self-hosting required.

Durable jobs, not fire-and-forget

Every long-running call runs on Cyclonic workers — survives restarts, cancellable, resumable, streams via SSE. Production-safe by default.

Multi-workspace from day one

One master API key, N client workspaces. Switch with a single `X-Workspace-Id` header. Per-client data isolation, bulk ops across all of them.

Demo mode out of the box

Every endpoint responds in shape-complete demo mode without auth, so you can build the agent before the user signs up.

FAQ

Common questions

Connect Your LLM to CitationBench MCP

One hosted MCP server gives your LLM agent full SEO and GEO visibility across every client workspace.

View MCP Docs