CitationBenchTalk to Sales

Introduction to CitationBench — SEO & GEO ops platform for agents

CitationBench is the SEO and GEO ops platform for teams running organic visibility at scale. Dashboard, REST API, and hosted MCP server — ~130 endpoints across five pillars.

CitationBench is the SEO and GEO ops platform for teams running organic visibility at scale. It's a product of Synscribe, built on the same infrastructure that powers Synscribe's internal content operations.

This site is the developer documentation. The product itself is available three ways:

  1. The dashboard at app.citationbench.com (visual; not the subject of these docs)
  2. The REST API at api.citationbench.com/v1/* (typed, OpenAPI 3.1)
  3. The hosted MCP server at mcp.citationbench.com/mcp (agentic; works with Claude Code, Cursor, Claude Desktop, Windsurf, ChatGPT Apps, and any MCP client)

All three are backed by the same handlers, the same auth, and the same audit log.

What you can do

Five pillars. Roughly 130 endpoints. Every endpoint is also available as an MCP tool.

PillarWhat it covers
AgentInvoke composable agents (bootstrap_brand, rank_monitor, link_hunter, ...), upload files for them to read, manage approval gates
ResearchKeyword research, SERP analysis, competitor backlinks, ICP generation, Reddit pain-point research, AI search citation tracking, content gap analysis
ProductionBlog posts, landing pages, content refiners, content evaluation, OG images, multi-platform publishing
IndexingGoogle Search Console indexing (auto-fires on publish), IndexNow submission to Bing/Yandex/partners
Link BuildingSERP outreach campaigns, competitor-backlink outreach, campaign management, full CRM for accounts/relationships/contacts/events

→ Master catalog: API · Inventory (every endpoint listed)

Two ways to start

If you're a developer

# 1. Create an API key in the dashboard.
# 2. Hit the keywords endpoint.

curl -X POST https://api.citationbench.com/v1/research/keyword \
  -H "Authorization: Bearer sk_live_***" \
  -H "X-Workspace-Id: ws_***" \
  -H "Content-Type: application/json" \
  -d '{ "seed": "project management software", "limit": 50 }'

Continue to Quickstart — REST →

If you're an agent builder

# 1. Create an API key in the dashboard.
# 2. Add the CitationBench MCP server to your client.

claude mcp add citationbench https://mcp.citationbench.com/mcp \
  --header "Authorization: Bearer sk_live_***" \
  --header "X-Workspace-Id: ws_***"

# 3. Ask Claude to run a tool.
# > "research keywords for acme.com, problem-solution intent"

Continue to Quickstart — MCP →

Core ideas

A short list before you go further. Each links to a concept doc.

  • Workspace. A single brand or domain. All keywords, content, and rank data are scoped to a workspace. Agencies have many workspaces under one master account.
  • Agent. A named, durable, multi-step process that uses CitationBench tools to deliver an outcome. Invoked the same way as any other endpoint; runs a graph of tool calls; can pause for approval. Eight built-in agents in v1.
  • Pillar. Every tool falls into one of Agent, Research, Production, Indexing, or Link Building.
  • Credits. The unit of usage. Every tool charges a transparent credit cost.
  • Job / invocation. Anything that takes more than a few seconds runs as a durable job. Jobs are idempotent, observable, and (optionally) approval-gated. Agent invocations are a specialized kind of job graph.
  • 2D keyword labeling. Every keyword has an intent and a relevance. It's the core data model that drives downstream tool defaults.
  • Content Refiners. Style / voice / quality templates applied to content post-generation.
  • Approval Workflows. Any step that touches the outside world can be gated by human approval.

What's in this docs site

  • Get started — quickstarts, auth, workspaces (you're here)
  • Concepts — explainers for the unique parts of CitationBench's model
  • API reference — every endpoint, every parameter, every error. Start at Inventory.
  • MCP tools — the catalog and per-client setup guides
  • Playbooks — end-to-end recipes for common outcomes. Start at Keyword research for a brand in 20 minutes.
  • Webhooks — outbound event subscriptions
  • Changelog — every release, every deprecation

A few principles you'll see throughout

  • Idempotency on every write. Pass an Idempotency-Key header; we'll never double-execute.
  • Approval gates on outside-world actions. Outreach emails, publishes, and indexing requests can be held until a human approves.
  • Workspace scoping by default. Pass X-Workspace-Id or use a workspace-scoped key. Agency master keys can target any sub-workspace.
  • Observable jobs. Every job has a parent–child graph, a tool-call timeline, a cost breakdown, and a replay link.
  • No silent retries. If a job fails, you get the error. We don't paper over it.
  • Naming guarantee. Every endpoint name is writable as the Google query for it — research.keyword, produce.blog_post, link_building.serp_outreach. If you can't query it, we rename it.

Stability

The /v1/ API is stable. Breaking changes follow a 12-month deprecation window with Sunset headers. Beta tools are prefixed /v1/beta/ and may change.

Get help

  • Status: status.citationbench.com
  • Discord: discord.citationbench.com
  • Email: support@citationbench.com
  • Enterprise / agency CS: through your account.

CitationBench · a product of Synscribe.

On this page