DigestSEO — AI Visibility MCP for SEO & GEO

Quick Install

Runs locally over stdio with your own API keys — all data stays on your machine (see Privacy Policy). Set at least one engine key (OPENAI_API_KEY, ANTHROPIC_API_KEY, GEMINI_API_KEY, PERPLEXITY_API_KEY, XAI_API_KEY, SERPAPI_API_KEY); engines without a key skip gracefully.

Runtime: Node.js 22.13+ (CI exercises Node 22 and 24).

Claude Desktop / any MCP client (npx):

{
  "mcpServers": {
    "digestseo": {
      "command": "npx",
      "args": ["-y", "@digestseo/mcp-geo"],
      "env": {
        "OPENAI_API_KEY": "sk-...",
        "GEMINI_API_KEY": "your_key_here"
      }
    }
  }
}

ChatGPT (remote MCP): ChatGPT does not connect directly to local STDIO MCP servers. For ChatGPT, use the self-hosted remote MCP setup below, or the OpenAI Secure MCP Tunnel setup for a server running on a local/private machine. The public geo-mcp.digestseo.com/mcp endpoint is not a turnkey no-key fresh-scan service.

Claude Code:

claude mcp add --transport stdio digestseo -s user --env GEMINI_API_KEY=your_key_here -- npx -y @digestseo/mcp-geo

Or install the same local MCP integration through this repository's owner-controlled Claude Code marketplace:

/plugin marketplace add AKzar1el/mcp-geo
/plugin install digestseo-geo@digestseo-mcp

The marketplace plugin uses the repository's .mcp.json to launch npx -y @digestseo/mcp-geo. Zero provider keys are enough for tool discovery; for engine-backed scans, make only the provider keys you want available to the Claude Code process. The direct claude mcp add command above remains the simplest option when you want to attach provider keys explicitly to the server configuration.

Codex CLI:

codex mcp add digestseo -- npx -y @digestseo/mcp-geo

The zero-key command is enough for tool discovery. Add only the provider keys you want with repeated --env NAME=VALUE options before the -- when engine-backed scans are needed.

Amp CLI:

amp mcp add digestseo -- npx -y @digestseo/mcp-geo

Amp runs this as a local STDIO MCP server. The zero-key command is enough for tool discovery; before engine-backed scans, make only the provider keys you want available to the Amp process or configure them in Amp's local MCP env settings instead of committing secrets. See Amp's current MCP guide.

OpenCode v2:

opencode mcp add digestseo --global -- npx -y @digestseo/mcp-geo

OpenCode v2 runs this as a local STDIO server. Omit --global for project-only configuration; add only the provider keys you want with --env NAME=VALUE before the --. Zero keys still allow MCP tool discovery.

Mistral Vibe Code: add mcp-geo to the user-level ~/.vibe/config.toml or project-level ./.vibe/config.toml:

[[mcp_servers]]
name = "digestseo"
transport = "stdio"
command = "npx"
args = ["-y", "@digestseo/mcp-geo"]

The zero-key entry is enough for tool discovery. For engine-backed scans, pass only the provider keys you want through Vibe's STDIO environment configuration or the environment inherited by Vibe instead of committing secrets. Use /mcp digestseo (or /mcp) in Vibe to verify the server and tools. See Mistral's current MCP server guide and Vibe configuration reference.

GitHub Copilot CLI:

copilot mcp add digestseo -- npx -y @digestseo/mcp-geo

The base install starts with zero provider keys so tool discovery works. Add only the engine keys you want with Copilot CLI's --env NAME=VALUE option before running scans.

Qoder CLI:

qoder mcp add digestseo -- npx -y @digestseo/mcp-geo
qoder mcp list

Qoder launches this as a local STDIO MCP server. The zero-key command is enough for tool discovery; make only the provider keys you want available to the Qoder process before engine-backed scans. If Qoder is already running, use /mcp reload to rediscover the server and tools. See Qoder's current MCP server guide and MCP reference.

Docker Agent: Docker Agent can launch local STDIO MCP servers directly from agent YAML. Add this toolset to the agent that should use mcp-geo:

toolsets:
  - type: mcp
    command: npx
    args: ["-y", "@digestseo/mcp-geo"]

The zero-key form is enough for tool discovery. For engine-backed scans, add only the provider keys you need under the toolset's env: map (Docker Agent supports ${env.NAME} expansion) instead of committing secret values. See Docker's current local MCP tool documentation.

goose: add mcp-geo as a local STDIO extension in ~/.config/goose/config.yaml (macOS/Linux) or %APPDATA%\Block\goose\config\config.yaml (Windows):

extensions:
  digestseo-geo:
    type: stdio
    name: digestseo-geo
    enabled: true
    cmd: npx
    args: ["-y", "@digestseo/mcp-geo"]
    timeout: 300

The zero-key extension is enough for tool discovery. Before engine-backed scans, configure only the provider environment variables you want for this extension through goose's extension settings / secret storage instead of putting raw API keys in the YAML file. The same server can also be added interactively with goose configure -> Add Extension -> Command-Line Extension. See goose's current extension setup and configuration reference.

GitLab Duo CLI: current GitLab Duo CLI releases can consume Claude-compatible plugin marketplaces directly. Register this repository and install the existing digestseo-geo plugin:

glab duo plugin marketplace add https://github.com/AKzar1el/mcp-geo.git
glab duo plugin install digestseo-geo@digestseo-mcp

The installed plugin loads the same local npx -y @digestseo/mcp-geo MCP server from .mcp.json. Zero provider keys allow discovery; make only the provider keys you want available to the GitLab Duo CLI process before engine-backed scans.

Factory Droid:

droid mcp add digestseo "npx -y @digestseo/mcp-geo"
droid mcp list

Droid runs this as a local STDIO MCP server. The zero-key install is enough for tool discovery; add only the provider keys you choose in Droid's user-level MCP configuration before engine-backed scans. Keep provider secrets out of project-level .factory/mcp.json files.

Amazon Q Developer (IDE): open the Q Developer chat panel ? Tools ? +, choose STDIO, name the server digestseo, set Command to npx, and add Arguments -y and @digestseo/mcp-geo. Add only the provider environment variables you want before running scans; zero keys still allow MCP tool discovery.

JetBrains AI Assistant (IDE): open Settings > Tools > AI Assistant > Model Context Protocol (MCP) > Add, choose STDIO, and use:

{
  "mcpServers": {
    "digestseo": {
      "command": "npx",
      "args": ["-y", "@digestseo/mcp-geo"]
    }
  }
}

JetBrains AI Assistant supports local STDIO and NPX MCP servers. The zero-key form is enough for tool discovery; before engine-backed scans, make only the provider keys you want available to the IDE process, or import an already-configured Claude MCP server.

JetBrains Air: this repository already ships the standard root .mcp.json that launches npx -y @digestseo/mcp-geo. In Air, open Settings > AI > MCP Servers, enable MCP support and Launch workspace MCP servers, then use the Workspace scope so Air reuses that checked-in file. The repository config contains no provider secrets and is sufficient for zero-key tool discovery. Engine-backed scans still require the selected provider keys in the local server process environment; keep them out of committed .mcp.json. See the JetBrains Air MCP server guide.

Visual Studio 2022 17.14+ / Visual Studio 2026: Visual Studio uses its own servers-shaped MCP configuration. Create %USERPROFILE%\.mcp.json for a user-wide install or <SOLUTIONDIR>\.mcp.json for one solution:

{
  "servers": {
    "digestseo": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@digestseo/mcp-geo"]
    }
  }
}

Open GitHub Copilot Chat in Agent mode and use the Tools menu to verify digestseo is available. Zero provider keys are enough for tool discovery; before engine-backed scans, make only the provider keys you want available to the Visual Studio process rather than committing secrets into the solution file. See Microsoft's current Visual Studio MCP setup.

Cursor:

Windsurf: open Manage MCPs → View raw config and add the local stdio package:

{
  "mcpServers": {
    "digestseo": {
      "command": "npx",
      "args": ["-y", "@digestseo/mcp-geo"],
      "env": {
        "OPENAI_API_KEY": "sk-..."
      }
    }
  }
}

Use only the provider keys you want; zero keys still allow MCP tool discovery.

Roo Code: open MCP Servers > Edit Global MCP, or create .roo/mcp.json for a project-scoped install, and add:

{
  "mcpServers": {
    "digestseo": {
      "command": "npx",
      "args": ["-y", "@digestseo/mcp-geo"]
    }
  }
}

Roo Code supports local STDIO MCP servers. The zero-key form is enough for tool discovery; add only the provider keys you want before scans. On Windows, if Roo cannot launch npx directly, use "command": "cmd" with "args": ["/c", "npx", "-y", "@digestseo/mcp-geo"], matching Roo's documented Windows fallback.

VS Code: Install the local stdio server in VS Code. The base install starts with no provider keys so tool discovery works; add only the API keys you want in your VS Code MCP user configuration before running scans.

LM Studio: Add the local stdio server to LM Studio. The base install starts with no provider keys so tool discovery works; add only the API keys you want in LM Studio's mcp.json before running scans.

Kiro: for a one-click local install, or open the user MCP config (~/.kiro/settings/mcp.json) / workspace config (.kiro/settings/mcp.json) and add the local stdio package manually:

{
  "mcpServers": {
    "digestseo": {
      "command": "npx",
      "args": ["-y", "@digestseo/mcp-geo"]
    }
  }
}

Kiro supports local STDIO MCP servers across its current surfaces. The zero-key form is enough for discovery; add only the provider environment variables you want before engine-backed scans. For remote use, self-host the Worker below and configure your own deployed /mcp URL instead. The public geo-mcp.digestseo.com/mcp endpoint is not a turnkey no-key fresh-scan service.

Claude Desktop extension (one-click): download the .mcpb bundle from the latest release and double-click it — Claude Desktop prompts for the API keys.

First run: ask your client to "track acme.com as brand acme, then refresh it" — track_brand creates the brand with generated prompts, refresh_brand runs the first scan, check_visibility shows the scores.

AI agents installing this server: follow llms-install.md. Prefer a remote server with cron auto-refresh? Self-host on Cloudflare Workers below.

mcp-geo is an open-source AI visibility tracker that measures how often your brand is cited by ChatGPT, Claude, Perplexity, Gemini, Grok, and Google AI Overviews. It's the GEO (Generative Engine Optimization) and AEO (Answer Engine Optimization) equivalent of Google Search Console — built as an MCP server so you can query your AI visibility data directly inside ChatGPT through a configured remote MCP app, Claude.ai, Claude Desktop, Claude Code, GitHub Copilot CLI, Cursor, Codex CLI, or any MCP-compatible client.

Canonical product page: DigestSEO mcp-geo — AI Visibility MCP Server

Engineering case study: DigestSEO MCP Suite — AI visibility, Search Console, web validation, and trend intelligence

Need a client-ready baseline without running the stack yourself? The mcp-geo AI Visibility Audit is EUR 99 one time: one brand, up to three competitors, 20 buyer-intent prompts, checks across up to five supported AI surfaces where configured providers return usable results, citation evidence, and a prioritized action memo. The open-source package remains free.

See proof first: Open the sample report generated through mcp-geo to see the output style and evidence depth before requesting the audit.

Ready to request it? Open a prefilled email with your brand/domain and up to three competitors. No subscription or sales call is required.

Payment handoff: After fit and scope are confirmed, I reply with the normal invoice/payment instructions.

Methodology: The same 20 buyer-intent prompts are run as a point-in-time diagnostic and reported per engine, with citation/source evidence where available. The audit is an observed snapshot, not a proprietary ranking promise or guaranteed forecast.

Want the protocol before buying? Read the AI Visibility Audit methodology, including scope, engine coverage, interpretation limits, and what the audit does not claim.

Prefer zero setup? Try the hosted version at digestseo.com — managed Cloudflare infra, no API keys to manage, multi-brand, scheduled refresh, web UI. Waitlist now open. Join waitlist →

What it produces

Connect via MCP, ask Claude "Run an AI visibility analysis on [my brand]", and within 90 seconds you get a strategist-quality memo grounded in real per-engine data:

View the full report including content gaps, engine recommendations, and synthesis →

Want to reproduce the same evidence-first structure with your own data? Use the reusable AI Visibility Audit report prompt.

The report above was generated by Claude through the digestseo-mcp MCP server. The conversation chained five hosted tools — visibility.check, visibility.compare, visibility.citations (Perplexity + Claude), and visibility.content_gaps — to produce a 4-engine analysis with citation excerpts and a 3-recommendation strategy memo.

What's New

[0.3.18] - September 21, 2026

• Grok visibility coverage: opt in with XAI_API_KEY to measure grounded Grok answers via xAI Web Search alongside ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews.

[0.3.17] - September 21, 2026

• Exact user-defined measurement sets: local MCP users can now call set_prompts to replace a brand's active prompts with 1-50 agreed buyer questions while preserving historical runs; repeated identical sets are a no-op.

[0.3.16] - September 21, 2026

• New-project Gemini compatibility: Gemini scans now default to gemini-3.1-flash-lite, avoiding the Gemini 2.5 access restriction Google applies to some new projects while preserving the same GenerateContent integration.

[0.3.15] - September 20, 2026

• Clearer provider setup failures: an explicit refresh request for an unconfigured engine now names the unavailable engine and explains how to recover instead of reporting that no engines at all are available.
• Broader local-client onboarding: copy-paste setup now covers Roo Code, Codex CLI, and OpenCode v2 in addition to the existing MCP clients.
• Reliable installed README links: npm-package readers are routed to durable GitHub URLs for documentation and report assets that are intentionally excluded from the tarball.

[0.3.14] - September 20, 2026

• Inspectable measurement inputs: local MCP users can now call list_prompts to review the exact active buyer-intent prompt set for a tracked brand without regenerating or changing it.
• Correct ChatGPT onboarding: ChatGPT guidance now uses a configured remote MCP server or OpenAI Secure MCP Tunnel for local/private servers instead of advertising unsupported direct local STDIO registration.

[0.3.13] - September 20, 2026

• More faithful citation evidence: get_citations now preserves the exact engine-native cited page URL, including path and query, when the provider returns one.
• Current Gemini CLI gallery metadata: the extension manifest now stays version-synchronized with the package, preventing stale gallery version labels after release.

[0.3.12] - September 20, 2026

• Better agent workflow guidance: local and hosted MCP initialize responses now include concise server-level instructions for the correct brand → refresh → visibility flow, asynchronous hosted refresh behavior, and unavailable-engine semantics.

[0.3.11] - September 19, 2026

• More reliable scheduled tracking: per-engine freshness and due-engine fan-out keep stale providers updating without needlessly rescanning fresh ones, while brands can choose daily or weekly cadence.
• Safer request handling: duplicate engine selections are deduplicated in hosted refresh and visibility reads, and hosted brand seeding rejects invalid refresh cadence values before persistence.
• Complete Claude Desktop metadata: the MCPB manifest now declares all nine fixed local tools, including track_brand, list_brands, and generate_prompts, matching the actual stdio server exposed after install.

[0.3.10] - September 19, 2026

• Safer BYOK refreshes: duplicate engine names are deduplicated before provider dispatch so one request cannot accidentally trigger duplicate scans or provider charges.
• Stronger MCP contracts: winning/losing prompt and content-gap structured outputs now publish concrete schemas instead of opaque records, improving client-side validation and agent interoperability.
• Better Registry installs: Official MCP Registry metadata now advertises all five supported provider API keys as optional secret configuration for the npm stdio package.

[0.3.9] - September 19, 2026

• Transparent snapshot freshness: visibility snapshots now expose each engine's own observation timestamp so older engine data cannot be mistaken for uniformly fresh results.
• Auditable history trends: visibility history now includes the usable-prompt denominator, brand-mention count, and observation time behind every per-engine score.

[0.3.8] - September 19, 2026

• More reliable audit evidence: competitor comparisons now honor their requested multi-day window across all usable runs and return exact mention counts; citation excerpts and provider-free content-gap fallbacks are grounded in the same underlying evidence.
• Broader local install reach: added Amazon Q Developer IDE setup for the existing local stdio package.

[0.3.7] - September 19, 2026

• Exact audit prompt count: prompt generation now persists exactly the requested number of unique prompts or leaves the existing prompt set untouched, protecting the paid audit's fixed 20-prompt scope.

[0.3.6] - September 18, 2026

• Grounded ChatGPT scans: live ChatGPT visibility uses web search, and domain-lookalike mention scoring was tightened.
• Broader local install reach: added Gemini CLI metadata plus one-click VS Code and LM Studio install paths.

[0.3.5] - September 17, 2026

• Provider/runtime correctness: migrated Perplexity to the Agent API and kept plugin installs on the local stdio package rather than the unconfigured public Worker.
• Trust/readiness: added owned privacy disclosure, production-only MCPB packaging, and transparent audit score formulas.

[0.3.4] - September 15, 2026

• Claude Desktop MCPB portability: the bundle no longer ships better-sqlite3 native binaries; local storage uses built-in node:sqlite on Node.js 22.13+.
• Registry accuracy: official metadata now advertises the npm stdio package only while the public hosted endpoint is not a turnkey configured fresh-scan service.

[0.3.3] - September 10, 2026

• Optional one-time audit: the open-source package stays free; teams that want a client-ready baseline can request the EUR 99 mcp-geo AI Visibility Audit from the CTA above.
• Lower-friction request path: the README now opens a prefilled, source-marked email, while the audit details remain available at https://geo-mcp.digestseo.com/audit.

[0.3.2] — July 27, 2026

• Published scoped package: @digestseo/mcp-geo with synchronized Worker, MCP Registry, and MCPB metadata.
• Hosted tool metadata: visibility.* namespaces with typed input/output schemas; local stdio tool names remain flat.
• Distribution and deployment: dedicated mcp-geo-db D1 configuration, Cursor and Claude Code plugin metadata, and patched production dependency pins.

[0.3.0] — July 2026

• Local stdio CLI on npm (npx -y @digestseo/mcp-geo): the same MCP tools backed by a local SQLite database (~/.digestseo/digestseo.sqlite) — no Cloudflare account needed. Engines run inline with your own API keys.
• Local brand-management tools (CLI only): track_brand, list_brands, generate_prompts. Workers deployments keep these behind the X-Seed-Secret-gated /admin/* routes.
• Runtime-agnostic core (src/core/) shared by the Worker and the CLI, with a Db contract implemented by D1 and better-sqlite3 adapters. All 0.2.1 accuracy and security fixes carry over to both runtimes.
• Distribution metadata: official MCP Registry server.json, MCPB desktop extension (.mcpb bundle), Dockerfile, llms-install.md for AI agents, release-publish workflow.

[0.2.1] — June 2026

• Optional CONNECT_SECRET gate on the OAuth flow. By default the OSS build auto-completes /authorize for any MCP client that knows your worker URL — anyone who finds the URL can connect and call visibility.refresh, spending your engine API credits. Set CONNECT_SECRET and the browser step of the connect flow now asks for it before issuing a token. See SECURITY.md.
• Accurate citation matching. Brand/competitor mentions now require word boundaries (acme no longer matches "acmeshop"), and linked-citation checks require the exact domain or a subdomain (notacme.com no longer counts as a link to acme.com).
• Per-brand aliases and exclude_terms. Aliases always count as a mention; exclude terms suppress the bare-word match on the brand name and domain root — so "Monday" the brand stops matching "monday" the weekday, while monday.com still counts. Apply migrations/0005_brand_alias_exclude.sql; existing brands behave exactly as before.
• visibility.history consistency. Partially-finished runs now count toward history (matching visibility.check's 0.2.0 behavior), and fully-failed runs no longer show up as fake zero scores.
• CI + unit tests. GitHub Actions runs tsc --noEmit plus a pure-function unit suite (npm run test:unit) covering mention matching, citation extraction, and score aggregation on every push.
• Docs now recommend OpenAI + Anthropic as the starting engine pair — Gemini capacity varies by model, project, and usage tier, so provider-specific quota variability could produce misleading first-run data as the documented cheapest path.
• Constant-time comparison for SEED_SECRET / CONNECT_SECRET.

[0.2.0] — May 2026

• Per-engine HTTP fan-out. /admin/run-live now creates one runs row per engine and self-fetches /admin/run-engine once per engine. Each engine runs in its own worker invocation with its own free-plan 50-subrequest budget — a single-invocation fan-out used to burst past the cap mid-run and lose half the rows.
• Service binding (env.SELF) dispatches the per-engine fan-out through Cloudflare's internal fabric instead of a public-URL fetch, dodging the "Worker called itself" guard (error 1042) that silently blocks the latter.
• Status column on prompt_responses (ok / failed / skipped) plus error_message. Failed engine calls used to write raw_response='ERROR: ...' rows that downstream scoring treated as real zero-mention hits; now they're explicitly excluded.
• FK-resistant inserts. /admin/run-engine INSERT OR IGNOREs its runs row before persisting — D1 is eventually consistent across edge regions, and the upstream INSERT INTO runs from /admin/run-live doesn't always replicate before the downstream engine call lands. The IGNORE makes the FK happy either way.
• Bulk D1 batch. Each engine collects its 20 prompt results in memory then flushes inserts + cache writes + the final UPDATE runs SET status='completed' in a single D1.batch() call. Drops the per-invocation subrequest count from ~89 to ~26.
• Relaxed visibility queries. getLatestCompletedRun anchors on EXISTS(ok rows) instead of status='completed', so partially-finished runs still surface their data in MCP tool output instead of silently disappearing.
• New admin route POST /admin/cleanup-failed-runs for one-shot deletion of legacy polluted rows after migrating to 0004.

[0.1.1] — May 2026

• Manual install is now the canonical path. The unreliable bash setup script was removed; SETUP.md is self-contained and copy-pasteable, with every interactive wrangler prompt documented inline.

[0.1.0] — May 2026

• Initial public release.
• 5-engine support: ChatGPT (gpt-4o-mini), Claude (claude-haiku-4-5), Perplexity (sonar), Gemini (gemini-2.5-flash-lite), and Google AI Overviews (via SerpAPI).
• 6 hosted MCP tools: visibility.check, visibility.history, visibility.compare, visibility.citations, visibility.content_gaps, visibility.refresh.
• Engines are opt-in based on which API keys you provide — set only the credentials you have, the rest skip gracefully.
• Cloudflare Cron Trigger that auto-refreshes tracked brands every 6h, respecting per-brand refresh_frequency (daily/weekly).
• D1-backed storage for brands, prompts, runs, citations, and a shared prompt cache.

What Can This Do?

• See which AI tools cite your brand and which don't — get a per-engine breakdown of who's citing you for buyer-intent queries.
• Track AI visibility weekly, automatically — the built-in Cron Trigger re-runs scans on the cadence you configure per brand.
• Compare your AI visibility to competitors — share-of-voice percentages, prompts you win, prompts they win.
• Find content gaps — Claude-Haiku-synthesized recommendations grounded in your actual losing prompts.
• Use it inside Claude.ai conversations — add the deployed Worker URL as a custom MCP connector and ask in natural language.
• Self-hosted on your own Cloudflare account — your API keys, your data, your cost ceiling. The free Workers + D1 tiers cover a single brand with daily refreshes.

See the example report above for what this looks like in practice.

Available Tools

The six analysis capabilities are shared across both transports, but the exposed MCP names are intentionally transport-specific: hosted/Worker connections use the visibility.* namespace, while the local stdio package uses flat names.

The local stdio CLI (npx, desktop extension, Docker) additionally provides brand management — on a Workers deployment the same operations live behind the X-Seed-Secret-gated /admin/* routes instead:

Getting Started

Step 1 — Get API keys

Engines are opt-in. Pick the ones you want; the rest skip silently.

• OpenAI — ChatGPT engine (gpt-5-search-api) with web search. OpenAI currently bills web search at $10 per 1,000 calls plus model token charges; see API pricing and API keys.
• Anthropic — Claude engine, plus prompt generation and content-gap analysis (both call Claude Haiku). ~€0.0002 per prompt. Free trial credits are usually enough to evaluate. console.anthropic.com
• Google AI Studio (Gemini) — Gemini engine (gemini-3.1-flash-lite). Google currently offers free-tier token usage for this model, while paid usage is token-priced. Rate limits vary by model, project, and usage tier, and Google says actual capacity can vary; check your project's active limits in AI Studio rather than relying on a fixed RPM/RPD assumption. See Gemini pricing and rate limits.
• Perplexity — Perplexity Sonar engine. ~€0.005-0.008 per prompt. Paid only. perplexity.ai/settings/api
• xAI — Grok engine (grok-4.6) with required Web Search grounding. xAI currently prices Web Search at $5 per 1,000 calls plus model tokens. console.x.ai · pricing
• SerpAPI — Google AI Overviews engine. ~€0.005 (free tier) / ~€0.0015 (volume) per prompt. Free tier covers 250 searches/month — enough for development. serpapi.com/dashboard

Recommended starting pair: OpenAI + Anthropic (Claude). OpenAI provides grounded ChatGPT visibility through web search and bills search calls plus model tokens; Anthropic also powers prompt generation and content-gap analysis. Review current provider pricing before estimating recurring scan cost. Add Gemini, Perplexity, Grok, or SerpAPI deliberately once you want more coverage; Gemini capacity varies by model, project, and usage tier, and Google AI Overviews often returns no result (scored as a zero), so leading with the cheapest path can skew your first run.

Step 2 — Deploy to your Cloudflare account

The deploy is 6 commands and takes about 5 minutes. See SETUP.md for the full walkthrough with explanations and troubleshooting, or follow the quick version below.

# 1. Install deps
npm install

# 2. Log in to Cloudflare
npx wrangler login

# 3. Copy the config template
cp wrangler.example.jsonc wrangler.jsonc

# 4. Create KV namespace + D1 database, paste each printed id into wrangler.jsonc
npx wrangler kv namespace create OAUTH_KV
npx wrangler d1 create mcp-geo-db

# 5. Set the required secret + at least one engine API key
#    Recommended starting pair — OpenAI uses web search plus model tokens; check current pricing:
npx wrangler secret put SEED_SECRET
npx wrangler secret put CONNECT_SECRET      # recommended — gates who can connect (see SECURITY.md)
npx wrangler secret put OPENAI_API_KEY      # ChatGPT engine
npx wrangler secret put ANTHROPIC_API_KEY   # Claude engine + prompt generation

# 6. Apply migrations and deploy
npx wrangler d1 migrations apply mcp-geo-db --remote
npx wrangler deploy

After deploying your own Worker, use that deployment's /mcp URL as the remote endpoint, for example:

https://YOUR-WORKER-NAME.YOUR-SUBDOMAIN.workers.dev/mcp

Use your configured Worker URL for directory or client integrations. The public geo-mcp.digestseo.com/mcp endpoint is not a no-key hosted substitute for a deployment with engine provider credentials.

Step 3 — Connect to your MCP client

After wrangler deploy finishes, you get a URL like https://digestseo-mcp.YOUR-SUBDOMAIN.workers.dev.

Claude.ai (web)

Settings → Connectors → Add custom connector. Paste:

https://YOUR-WORKER-NAME.YOUR-SUBDOMAIN.workers.dev/mcp

Complete the OAuth handshake. The connector turns green when ready.

ChatGPT (remote MCP)

ChatGPT custom MCP apps connect to remote MCP servers, so use the /mcp URL from your configured Worker deployment above. In ChatGPT, enable Developer Mode/custom apps for your workspace and add that remote MCP URL. Availability depends on your ChatGPT plan and workspace admin policy; OpenAI's current MCP support does not require special search or fetch tool names.

https://YOUR-WORKER-NAME.YOUR-SUBDOMAIN.workers.dev/mcp

ChatGPT (local/private via OpenAI Secure MCP Tunnel)

OpenAI Secure MCP Tunnel is the supported bridge when you want ChatGPT to use the local stdio package without exposing it as a public HTTPS server. Create a tunnel in OpenAI Platform first, then keep tunnel-client running on the same machine that launches mcp-geo. You need a tunnel ID, a tunnel runtime API key, and ChatGPT developer-mode/tunnel permissions for the target workspace.

Make whichever provider keys you want to use available to the mcp-geo process, then initialize a tunnel profile with the local package command:

tunnel-client init --sample sample_mcp_stdio_local --profile digestseo --tunnel-id tunnel_0123456789abcdef0123456789abcdef --mcp-command "npx -y @digestseo/mcp-geo"
tunnel-client doctor --profile digestseo --explain
tunnel-client run --profile digestseo

While tunnel-client run is healthy, create a developer-mode app in ChatGPT, choose Tunnel as the connection type, and select that tunnel. This path is for private/local use; it does not publish mcp-geo as a public ChatGPT app. Follow OpenAI's current Secure MCP Tunnel guide for tunnel creation, permissions, downloads, and troubleshooting.

Claude Code

claude mcp add --transport http digestseo https://YOUR-WORKER-NAME.YOUR-SUBDOMAIN.workers.dev/mcp

Then run /mcp inside Claude Code to complete the OAuth handshake in your browser.

Claude Desktop

Edit your Claude Desktop config:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "digestseo": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://YOUR-WORKER-NAME.YOUR-SUBDOMAIN.workers.dev/mcp"
      ]
    }
  }
}

Restart Claude Desktop after editing.

Cursor

Edit ~/.cursor/mcp.json:

{
  "mcpServers": {
    "digestseo": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://YOUR-WORKER-NAME.YOUR-SUBDOMAIN.workers.dev/mcp"
      ]
    }
  }
}

Restart Cursor.

Codex CLI

Add to ~/.codex/config.toml:

[mcp_servers.digestseo]
command = "npx"
args = [
  "-y",
  "mcp-remote",
  "https://YOUR-WORKER-NAME.YOUR-SUBDOMAIN.workers.dev/mcp",
]

Environment Variables Reference

All values are set via wrangler secret put VAR in production or .dev.vars locally. None are stored in wrangler.jsonc.

Architecture

flowchart LR
    C["MCP client<br/>(Claude.ai / Claude Code / Cursor / ...)"] -- "MCP over HTTP + OAuth" --> W["Cloudflare Worker<br/>digestseo-mcp"]
    CRON["Cron Trigger<br/>every 6h"] --> W
    W --> DO["GeoMcpAgent<br/>(Durable Object, 6 MCP tools)"]
    W -- "one self-fetch per engine<br/>via SELF service binding" --> RE["/admin/run-engine<br/>(own invocation per engine)"]
    RE --> E1["OpenAI"]
    RE --> E2["Anthropic"]
    RE --> E3["Gemini"]
    RE --> E4["Perplexity"]
    RE --> E5["xAI<br/>(Grok)"]
    RE --> E6["SerpAPI<br/>(AI Overviews)"]
    RE --> DB[("D1<br/>brands / prompts / runs /<br/>responses / cache")]
    DO --> DB

Each engine runs in its own Worker invocation with its own free-plan 50-subrequest budget; results are flushed in a single D1.batch() per engine. The whole system fits the Cloudflare free tier for a single brand on a daily cadence.

Security

• /admin/* is gated by SEED_SECRET (constant-time compared).
• /mcp requires OAuth; set CONNECT_SECRET so only people with the secret can complete the connect flow — strongly recommended whenever your worker URL is shared anywhere, since connected clients can call visibility.refresh and spend your engine API credits.
• All engine keys live in Cloudflare's encrypted secret store; all data stays in your own D1 database.

Full details and vulnerability reporting: SECURITY.md.

Sample Prompts

The example report above was generated by the first prompt below.

Once the connector is live in Claude.ai (or any MCP client), try:

Hosted Version

If you'd rather not run your own Cloudflare account, manage API keys, or pay individual engine bills, the hosted version of DigestSEO runs the same MCP server on managed infrastructure with multi-brand support, scheduled refresh, a web UI, and consolidated billing. Waitlist now open — join at digestseo.com.

Troubleshooting

• Worker deploys but tools return empty data — at least one engine API key is missing. Check wrangler secret list and add the keys you intend to use. Engines without keys are silently skipped, which can leave visibility.check with no data.
• no engines available error in logs — no engine API keys are set at all. Set at least one of OPENAI_API_KEY, ANTHROPIC_API_KEY, GEMINI_API_KEY, PERPLEXITY_API_KEY, XAI_API_KEY, SERPAPI_API_KEY.
• D1 migration fails — make sure you've run npx wrangler d1 migrations apply mcp-geo-db --remote (and also --local for wrangler dev). For ad-hoc fixes, npx wrangler d1 execute mcp-geo-db --remote --file=migrations/0001_initial.sql.
• Custom MCP connector in Claude.ai not connecting — the URL must end in /mcp. The OAuth handshake auto-completes in the OSS build (single dev user); if you set CONNECT_SECRET, the browser step shows a one-field form — enter the secret you set during deploy. If it loops, clear the connector and re-add it. Double-check the Worker is publicly reachable (curl https://YOUR-WORKER-NAME.YOUR-SUBDOMAIN.workers.dev/healthz should return ok).
• Cron not firing — check the Cloudflare dashboard at Workers & Pages → digestseo-mcp → Settings → Triggers. The "Cron Triggers" section should list 0 */6 * * *. If it's missing, run npx wrangler deploy again — the trigger is registered on deploy. The handler also only dispatches engines for brands whose refresh_frequency cadence has elapsed, so a freshly-seeded brand might not fire on the next 6h boundary.
• 401 unauthorized from /admin/* — X-Seed-Secret header is missing or doesn't match the deployed SEED_SECRET. Re-run npx wrangler secret put SEED_SECRET and update your .env.test.
• Worker returns 404 on self-fetch / error code 1042 — the services binding in wrangler.jsonc is missing or the service name doesn't match the worker's name field. /admin/run-live self-fetches /admin/run-engine via env.SELF (a Cloudflare service binding) precisely because a public-URL fetch back to your own workers.dev hostname is blocked by Cloudflare's "Worker called itself" guard. Confirm the wrangler.jsonc you deployed contains "services": [{ "binding": "SELF", "service": "<your-worker-name>" }] with the same name you set in the top-level "name" field. After fixing, npx wrangler deploy and re-run.
• Gemini rate limit (HTTP 429) on prompts — Gemini API limits vary by model, project, and usage tier, and Google notes that actual capacity can vary. Check the project's current limits in Google AI Studio and Google's rate-limit documentation. When a request is rate-limited, mcp-geo records that engine row as failed and excludes it from successful scoring. Google recommends waiting and retrying after a short period or reducing request rate; if the limit is consistently too low for your scan cadence, consider an appropriate paid tier rather than assuming a universal free-tier RPM/RPD quota.
• FOREIGN KEY constraint failed in wrangler tail during /admin/run-engine — the handler defensively INSERT OR IGNOREs the runs row before persisting prompt responses. This is an idempotency/FK guard for independently dispatched engine work, so you should not see this on the 0.2.0+ build; if you do, confirm you've deployed the latest src/index.ts (grep -n "INSERT OR IGNORE INTO runs" src/index.ts should match).

Contributing

Issues and PRs welcome. See CONTRIBUTING.md for the short version.

Privacy Policy

Full policy for the local package and Claude Desktop extension: https://geo-mcp.digestseo.com/privacy

When you run digestseo-mcp locally (npx, the desktop extension, or Docker), all of your data — brands, prompts, runs, responses, and the response cache — stays on your machine in a local SQLite database at ~/.digestseo/digestseo.sqlite (override with DIGESTSEO_DB_PATH). The scan prompts are sent only to the AI providers whose API keys you configure (OpenAI, Anthropic, Google, Perplexity, xAI, and/or SerpAPI); their handling of that traffic is governed by their respective privacy policies. Nothing is ever sent to the author of this project: no telemetry, no analytics, no account.

Data use and storage: Local brand configuration, prompts, scan runs, responses, and cached responses are used only to provide the MCP server features you invoke. They remain in the local SQLite database described above; this project does not operate an account service or collect telemetry.

Third-party processing: Prompt and scan traffic is sent only to the AI providers you explicitly configure. Those providers process and retain that traffic under their own privacy policies; the project author does not receive copies of it.

Retention and deletion: Local data remains on your machine until you delete the SQLite database (or the custom DIGESTSEO_DB_PATH you configured). Removing that local database removes mcp-geo's stored local history and cache. Provider-side retention is controlled by each configured provider.

Contact: Privacy questions about mcp-geo can be sent to info@tomiseregi.si.

License

MIT.

Built and maintained by Tomi Šeregi.

Changelog

See CHANGELOG.md for the full version history.

[0.3.2] — July 27, 2026

• Published @digestseo/mcp-geo with synchronized Worker, MCP Registry, and MCPB metadata.
• Hosted visibility.* tool namespaces with typed input/output schemas; local stdio names remain flat.
• Dedicated mcp-geo-db D1 configuration and Cursor/Claude Code plugin metadata.
• Patched production dependency pins.

[0.3.0] — July 2026

• Local stdio CLI on npm (npx -y @digestseo/mcp-geo) with SQLite storage and inline engine runs.
• Local brand-management tools: track_brand, list_brands, generate_prompts.
• Runtime-agnostic core shared by Worker and CLI; D1 + better-sqlite3 Db adapters.
• MCP Registry server.json, MCPB desktop extension, Dockerfile, llms-install.md.

[0.2.1] — June 2026

• Optional CONNECT_SECRET gate on the OAuth connect flow.
• Word-boundary brand/competitor matching; exact-domain-or-subdomain linked-citation checks.
• Per-brand aliases and exclude_terms (migration 0005) for homograph brands like Monday/Notion.
• visibility.history includes partial runs and drops fully-failed runs.
• CI workflow (typecheck + unit tests) and a pure-function unit test suite.
• Docs recommend OpenAI + Anthropic as the starting engine pair.
• Constant-time secret comparison.

[0.2.0] — May 2026

• Per-engine HTTP fan-out via env.SELF service binding (one worker invocation per engine, dodges Cloudflare's 1042 self-call guard).
• status + error_message columns on prompt_responses — failed engine calls are now explicit rows, no more ERROR: strings in raw_response.
• INSERT OR IGNORE on the runs row inside /admin/run-engine (handles D1 cross-region replication lag without dropping prompt_responses to FK violations).
• Bulk D1 batch in each engine's runLive (~26 subrequests/invocation instead of ~89; full 20-prompt runs now fit under the free-plan cap).
• getLatestCompletedRun anchored on EXISTS(ok rows); partially-finished runs still show their data.
• New POST /admin/cleanup-failed-runs admin route.

[0.1.1] — May 2026

• Removed the unreliable bash setup script. Manual install via SETUP.md is now the canonical path.

[0.1.0] — May 2026

• Initial public release.
• 5-engine support: ChatGPT, Claude, Perplexity, Gemini, Google AI Overviews.
• 6 MCP tools.
• Engines opt-in based on which API keys you provide.
• Cloudflare Cron Trigger for auto-refresh.