PendiumDocs

Overview

Pendium documentation — AI visibility, content workflows, agent-site hosting, and the MCP + REST integration surface.

Pendium is a GEO/AEO platform: it monitors and improves how AI assistants (ChatGPT, Claude, Gemini, Google AI Overviews) perceive and recommend your brand. These docs cover the product concepts, your hosted agent site, and the programmatic integration surface.

New here? Start with the Quickstart.

What's in these docs

Connect your AI agent

Pendium gives you three ways to access AI visibility data, content tools, and brand intelligence:

MCP Server — The recommended path for AI agents. Your agent discovers tools automatically and gets structured responses with context. Works with Claude.ai, ChatGPT, Claude Code, Cursor, Windsurf, and any MCP-compatible client.

https://pendium.ai/api/mcp

Claude.ai / ChatGPT Connector — Add Pendium as a connector directly in Claude.ai or as an app in ChatGPT. No API key needed — you sign in with your Pendium account via OAuth.

REST API — Direct HTTP access for custom integrations, scripts, and environments that don't support MCP. Same data, standard curl/fetch interface.

https://pendium.ai/api/visibility

What you can do

35 MCP tools and 9 REST API endpoints across six categories:

  • Visibility — Trigger scans, poll results, get full reports, track score history, manage action items, inspect cited sources, run deep URL analyses and shareable preview reports, and look up any brand's public score.
  • Content Strategy — Access content goals, brand voice guidelines, factsheets, and analyzed URLs.
  • Content Workflows — Turn a seed into a batch of blog and social posts using brand voice, knowledge, and goals.
  • Knowledge — Search a brand's knowledge base semantically and list ingested sources.
  • Configuration — Manage topics, queries, and buyer personas (create, update, delete), and view account info.
  • Utility — Send feedback to the Pendium team from an agent session.

The MCP server exposes all 35 tools. The REST API covers the core scanning, reporting, and configuration endpoints.

Quick start

Pick your platform and follow the setup instructions.

Connect Claude.ai or ChatGPT

These platforms connect via OAuth — no API key needed. You sign in with your Pendium account and the platform gets access automatically.

Claude.ai (web and desktop app):

  1. Go to Settings → Connectors (or Customize → Connectors)
  2. Click Add custom connector
  3. Enter the MCP server URL: https://pendium.ai/api/mcp
  4. Click Connect — you'll be redirected to sign in with your Pendium account
  5. After signing in, Claude.ai has access to all 35 Pendium tools

ChatGPT:

  1. Go to Settings → Connectors (or click your profile → Settings)
  2. Click Add App (or New App)
  3. Set the MCP Server URL to: https://pendium.ai/api/mcp
  4. Set Authentication to OAuth
  5. Click Advanced settings — ChatGPT will auto-discover the OAuth configuration
  6. Under Client registration, select Dynamic Client Registration (DCR)
  7. Check the consent box and click Create
  8. Click Connect — you'll be redirected to sign in with your Pendium account

After connecting, ask Claude or ChatGPT something like:

"Use Pendium to scan the AI visibility for my brand and tell me how AI platforms perceive it."

Connect a coding agent

For Claude Code, Cursor, Windsurf, and other MCP-compatible coding agents, add the MCP server URL. Your agent handles authentication automatically via OAuth — when you first connect, you'll be prompted to sign in with your Pendium account.

1. Add the MCP server

Any agent (universal) — works with Claude Code, Cursor, Windsurf, Claude Desktop, and any MCP-compatible agent. Add this JSON to your agent's MCP config file:

{
  "mcpServers": {
    "pendium": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://pendium.ai/api/mcp"]
    }
  }
}

Config file locations:

  • Cursor: ~/.cursor/mcp.json
  • Claude Code: ~/.claude.json
  • Windsurf: ~/.codeium/windsurf/mcp_config.json
  • Claude Desktop: ~/Library/Application Support/Claude/claude_desktop_config.json

After adding, restart your agent (reload window in Cursor, restart CLI for Claude Code). You'll be prompted to sign in with your Pendium account on first use.

Platform-specific alternatives

Claude Code — one command:

claude mcp add --transport http pendium https://pendium.ai/api/mcp

Cursor (native url format in ~/.cursor/mcp.json):

{
  "mcpServers": {
    "pendium": {
      "url": "https://pendium.ai/api/mcp"
    }
  }
}

Windsurf (native format in ~/.codeium/windsurf/mcp_config.json):

{
  "mcpServers": {
    "pendium": {
      "serverUrl": "https://pendium.ai/api/mcp"
    }
  }
}

2. Start using it

Ask your agent something like:

"Scan the AI visibility for https://example.com and tell me how AI platforms perceive the brand."

The agent will use scan_visibility to trigger a scan, then poll with get_scan_status until results are ready.

Transport

The server uses Streamable HTTP transport, the standard MCP remote transport. Each request is stateless — no session management required. The server accepts GET, POST, and DELETE requests.

Agent prompt

Copy this into your agent's system prompt or project instructions to give it an overview of every Pendium tool. This isn't required — MCP tool descriptions are auto-discovered — but it helps agents pick the right tool faster.

Pendium tracks how ChatGPT, Claude, Gemini, and AI Overviews perceive,
mention, and recommend brands. Use it to ground your content strategy
in real data about what AI platforms actually say.

35 TOOLS:

  VISIBILITY MONITORING
  scan_visibility      — trigger a scan (use mode: "batch", maxQueries: 30)
  get_scan_status      — poll every 30-60s until scanAvailable=true
  get_report           — full report with raw LLM responses and recommendations
  get_scan_history     — historical scores to track trends over time
  get_recommendations  — prioritized actions to improve AI visibility
  complete_action      — mark an action as done after you've addressed it
  get_cited_sources    — URLs AI platforms cite when discussing the brand
  analyze_url          — deep competitive analysis of a URL (async; poll get_analyzed_url)
  get_analyzed_url     — poll/read an analyze_url result
  trigger_visibility_scan_preview — lightweight shareable brand-page report (async)
  get_visibility_scan_preview     — poll/read a preview report by uuid or slug
  lookup_brand_score   — public brand score lookup (no auth needed)

  CONTENT STRATEGY
  get_goals            — strategic business goals that frame content work
  get_brand_voice      — brand voice guidelines, tone, and company profile
  get_factsheet        — comprehensive brand overview and key facts
  list_analyzed_urls   — analyzed URLs in the pipeline

  CONTENT WORKFLOWS
  list_workflows       — list active content workflows for an agent
  create_workflow      — create a workflow from a seed (goal/trend/url_list/notes/…)
  add_workflow_ideas   — add specific post-idea rows (no LLM cost)
  generate_workflow_ideas — LLM-draft idea rows from the seed (paid)
  run_workflow_pipeline   — generate content for every pending row (paid)

  KNOWLEDGE BASE
  query_knowledge      — semantic search over brand's knowledge base
  list_knowledge_sources — view ingested content sources

  CONFIGURATION
  list_visibility_topics     — view current topics and queries
  add_visibility_topic       — add topics with search queries
  update_visibility_topic    — rename/edit/disable a topic
  delete_visibility_topic    — permanently delete a topic and its queries
  update_visibility_query    — edit a query's text, reach level, or active state
  delete_visibility_query    — permanently delete a query
  list_visibility_personas   — view buyer personas
  add_visibility_persona     — add personas (AI-generated from a prompt, or manual)
  update_visibility_persona  — edit or disable a persona
  delete_visibility_persona  — permanently delete a persona
  get_account                — check credits, plan, and list of agents

  UTILITY
  send_feedback        — report bugs, friction, or ideas to the Pendium team

Every response includes a _pendium field with context and next steps.

Claude Skill (optional)

A Claude Skill is a small file that teaches your agent workflow patterns — which tools to call, in what order, and what to avoid. Drop it in and your agent skips the learning curve.

Download SKILL.md

After downloading, place the file in your agent's skills folder:

  • Claude Code: ~/.claude/skills/pendium/SKILL.md
  • Cursor: ~/.cursor/skills/pendium/SKILL.md

The MCP works without the skill. It just makes your agent faster at picking the right tool for the job.

Next steps

MCP Tools:

REST API:

Next

Quickstart

On this page