Documentation
An AI agent as your team's strategist via MCP
Connect Claude, ChatGPT, Cursor, or Claude Code — and the team gains a strategist with direct access to the full monitoring picture. A guide to setup + examples: root-cause analysis of gaps, article briefs, reports — all right in chat
Open in dashboardAn AI agent as your team's strategist
With MCP you add a strategist to the team — one who works with monitoring data at a depth the UI can't reach. They read full AI responses, break down every cited source, compare with competitors, iterate on hypotheses, and assemble strategy and briefs. Ask "why does AI answer this way," "what should the article look like to make this prompt," "give me a quarterly content roadmap" — and get a full analytical breakdown, not a rehash of dashboard numbers. That strategist can be Claude, ChatGPT, Cursor, or any other MCP-compatible AI. Technically, MCP (Model Context Protocol) is the standard that gives the agent this access.
- Strategist clients: Claude Desktop, ChatGPT (via Custom Connectors), Cursor, Claude Code, Codex, or any other MCP-compatible client
- Connect via OAuth (pick the brand on the GEO Scout consent screen) or via a Personal Access Token (PAT) for scripts and self-hosted clients
- Read-only — the strategist reads and analyzes the data but cannot change anything in the system
- The strategist has access to everything the dashboard does, plus the full text of AI responses: metrics, sources, competitors, prompts, citations, actions, and content plans
An AI strategist isn't a "lightweight alternative" to the dashboard. It's a layer of deep work on top of the data: what a senior marketer or external consultant typically does over days, gets done in a single evening's conversation. Dashboard shows cuts, Command Center sets priorities, AI strategist does the analytics and strategy.
Claude Desktop
Claude Desktop speaks the stdio MCP protocol, so it connects to the HTTP server through the mcp-remote bridge (an npm package). npx installs it automatically — you just need Node.js 18+ on your machine.
- Open `claude_desktop_config.json` (on macOS: `~/Library/Application Support/Claude/`, on Windows: `%APPDATA%\Claude\`). If the file doesn't exist, create it
- Add an `mcpServers` block with the command `npx -y mcp-remote https://geoscout.pro/api/mcp` (the exact JSON snippet is on the /mcp page in the dashboard, in the Claude Desktop tab)
- Save the file and fully restart Claude Desktop. On first launch npx downloads mcp-remote and opens a browser for OAuth authorization
- On the GEO Scout consent screen pick the brand the agent should access and confirm. Multiple brands can be connected via separate MCP servers in the same config
- An MCP server indicator will appear at the bottom of the Claude Desktop chat window — that means the connection is live
For PAT-based access add `--header "Authorization: Bearer gs_<your-token>"` to args. On Windows there's a known mcp-remote quirk with space escaping — set the token via an `AUTH_HEADER` env var (value `Bearer gs_...`) and reference it in args as `Authorization:$AUTH_HEADER`. Exact snippets are on the /mcp page.
ChatGPT (Custom Connectors)
ChatGPT supports remote MCP servers through Custom Connectors on paid plans (Plus, Pro, Team, Enterprise). Set-up is fully UI-driven — no config files.
- Open chatgpt.com → Settings → Connectors → "Add custom connector"
- Paste the URL `https://geoscout.pro/api/mcp` and click "Add"
- ChatGPT auto-discovers the OAuth metadata and opens the GEO Scout consent screen — pick a brand and approve
- The connector activates immediately. In a new chat attach it via the paperclip and ask about your brand in natural language
Cursor
Cursor connects MCP servers through the "Connect to a custom MCP" UI form — no JSON files to edit. Works as a global connection or project-only.
- Open Cursor → Settings → MCP servers → "+ Add server"
- Fill in the dialog: Name = `geoscout`, Transport = `Streamable HTTP`, URL = `https://geoscout.pro/api/mcp`
- Save — Cursor opens a browser for OAuth and you pick a brand on the GEO Scout consent screen
- For PAT-based connection instead of OAuth: in the Headers block click "+ Add header", Key = `Authorization`, Value = `Bearer gs_<your-token>`
- After saving the server appears in the MCP servers list — the tools icon will be available in Composer and Chat
To connect the server only for one project, drop a `.cursor/mcp.json` file with the same config in the repo root — Cursor picks it up automatically and the connection won't leak into other projects.
Claude Code CLI
Claude Code — a one-line connect via `claude mcp add`. The OAuth flow runs in the browser, and the config is saved in the project.
- In the project root run: `claude mcp add --transport http geoscout https://geoscout.pro/api/mcp`
- Claude Code opens a browser for OAuth automatically. The `--transport http` flag is required for streamable HTTP MCP
- On the GEO Scout consent screen pick a brand and approve
- The connection is saved into the project's `.claude/` folder. For PAT add `--header "Authorization: Bearer gs_<token>"`
Your first question — where to start
Once connected, try a light question — the agent figures out which of the available tools to use. Ask in plain language; you don't need to know any API method names.
- "Give me the brand brief for the past week" — the agent returns coverage, Share of Voice, Citation Share, top gaps, and quick-win actions
- "Show me the top 10 sources AI cites about my brand" — a map of reference domains with categories (media, marketplaces, forums, competitors)
- "Which prompts are competitors beating us on the most?" — list of gap-prompts with deltas and the top competitor per prompt
- "Pull the top 3 negative responses from this week" — AI responses with negative sentiment plus context and provider
Don't try to guess tool names — describe the task. The agent picks from 24 available methods and chains them as needed. If a follow-up confuses the agent, rephrase: "Based on the monitoring data for the month, …"
Example: gap analysis down to root cause
A real workflow for an e-commerce brand: from spotting a gap to understanding why AI cites competitors. Each step is just a question in chat, all in one conversation.
- Step 1: "Where are competitors beating us the most?" — agent finds the top 5 prompts with the largest gap
- Step 2: "Take prompt #1 and tell me who AI cites and why" — detailed source breakdown and named competitors
- Step 3: agent returns: which domains are cited (e.g. retail.ru via a competitor case study), which competitors are mentioned in text, what content they use
- Step 4: "Why doesn't our brand show up there?" — agent compares your data with the competitor's and points at missing signals (no category page, no top-tier media mention, no JSON-LD markup)
- Step 5: all of this — in a single chat conversation. No need to open the dashboard, switch filters, or stitch findings by hand
The strength of MCP is exactly this — asking "why" and "what if." The dashboard shows "how it is now," MCP lets you dig for causes and find non-obvious connections. One conversation with the agent typically replaces 1–2 hours of manual dashboard work.
Example: drafting an article brief
Once a gap is understood, ask the agent to produce an article brief right away. It has everything it needs: the gap-prompt, competitor-cited sources, key entities from AI responses, your brand facts and certifications.
- Prompt: "Based on this gap, draft an article brief: title, URL, target key entities, schema markup"
- The agent will pull: the gap-prompt text, the list of competitor-cited domains, topical entities from 50+ AI responses, and your brand profile
- Output: title and URL for the publication, target key entities (must-mention items), JSON-LD blocks (Article + FAQPage), reference formats from retail.ru / vc.ru / Habr
- The brief can be saved to Notion/Linear/Jira via your AI's integrations (Claude — Tool Use, ChatGPT — connectors, Cursor — copy-paste)
- After publishing — ask the agent to track the impact 30 days later: "In a month, check whether this URL appears in AI responses and how Citation Share for the topic changed"
Brief quality depends on how you ask. Good templates: "give me 5 key entities," "suggest schema blocks," "find 3 facts about my brand from monitoring that would strengthen citation." The more specific the request, the denser the result.
Example: monthly / weekly report
Replace manual report-stitching with one command to the agent. It collects the numbers, the dynamics, the main shifts, and produces text you can send straight to a stakeholder or into a team chat.
- Prompt: "Prepare a monthly brand report — metric dynamics, top providers, major shifts, recommendations for next month"
- The agent calls get_brand_metrics, get_brand_metrics_timeseries, list_competitors, get_competitive_gaps and compares periods (month-over-month or week-over-week)
- Output: a structured text report with numbers, provider breakdown, shift explanations, and priority suggestions
- Follow up with refinements: "give me 3 hypotheses for the SoV drop on Perplexity," "propose 5 actions for next month based on current gaps"
What data is available to the agent
Through MCP the agent gets 24 read-only tools that cover every dashboard cut. You don't need to memorize them — the agent picks the right tool for each question.
- Metrics: coverage, Share of Voice, Citation Share, URL Mention Rate — by period, providers, prompt clusters, with period-over-period deltas
- Sources: cited domains with categories (media, catalogs, marketplaces, forums), individual URLs, competitor comparison, uncited pages on your site
- Competitors: list of active competitors, head-to-head on any metric, competitive gaps by cluster, mention dynamics
- Prompts and responses: top gap-prompts, per-prompt details with full AI responses, sentiment, scores, citations, mentioned competitors
- Actions and content plans: Command Center tasks, content plans with article structure, published pieces and their measured impact
The full list of available tools with descriptions is on the /mcp page in the dashboard. If you need a tool or data cut that isn't available yet — let support know, the MCP toolset is actively expanding.