# Bloomiro MCP (visibility intelligence in your workflow) Canonical URL: https://bloomiro.garden/mcp Markdown URL: https://bloomiro.garden/mcp.md Product updates: [Bloomiro Changelog](https://bloomiro.garden/changelog.md) Model Context Protocol (MCP) connects Cursor, Claude, ChatGPT, Codex, and VS Code to your live Bloomiro project. Your AI can use Google Search Console, AI mention checks, community citation sources, competitor gaps, page scans, and prioritized tasks from Bloomiro's proprietary Optimization and Action Layer, without leaving your workflow. ## MCP endpoint Use https://api.bloomiro.com/mcp with OAuth or a project API key from the Bloomiro dashboard. ## Agent usage 1. Call `get_project_context` first to resolve the site and check available integrations. 2. Pass the returned `site_id` to subsequent tool calls. 3. Call `get_mcp_guide` only when a request, field, or next step is genuinely ambiguous. 4. Use the exact tool names in the public catalog below. ## Connect 1. **Create a Bloomiro account.** Sign up, add your site, and connect Google Search Console if you use it. 2. **Authorize MCP.** Open the connect flow, sign in, pick a project, and approve access for your MCP client. 3. **Start chatting in your workflow.** Ask what to improve next, where competitors win, or which Action Center task to finish. Your AI can read and update the same shared Kanban you use in Actions. MCP is included on every plan (Starter, Builder, and Growth). It is not an add-on. Plans differ by sites, seats, and credits, not by Use in your AI access. The $BLOOMIRO access also includes it. ## Supported clients - Claude - ChatGPT - Perplexity - Cursor - VS Code - Claude Code - ChatGPT app (Codex) - Stdio ## What you can ask - **Fix SEO while you ship:** Pull live page scan issues, GSC queries, and shared Action Center tasks into Cursor or Claude Code. Use the findings to improve work your team has chosen to act on. - **Research competitors in chat:** Ask your assistant to compare your visibility vs competitors using real competitor scan data from your Bloomiro project, then decide what to improve next. - **Track AI mentions on buyer prompts:** Run AI presence checks from ChatGPT or Claude. See whether your brand shows up on the prompts your buyers actually ask. - **Inspect community pages AI already cites:** Ask which Wikipedia, Reddit, YouTube, LinkedIn, or review pages show up in answers, then open the exact URLs and prompt IDs from get_community_sources. - **Turn gaps into actions:** Keep agreed SEO and AI visibility work moving on the same Kanban in Actions chat, the dashboard, and MCP. Improve an existing task first. Create another only for a distinct, actionable outcome. ## Public MCP tool catalog (29 tools) Each tool is listed once. Optional integrations return setup guidance when a connection is required. ### Context and guidance Resolve the site once and get compact help only when needed. - `get_project_context`: Resolve site_id, domain, integrations, permissions, and data freshness. - `get_mcp_guide`: Compact routing help for ambiguous requests and unfamiliar fields. ### Search, SEO, and GEO Read Google data, audit evidence, technical checks, or start one typed analysis. - `get_search_performance`: GSC queries, pages, clicks, impressions, CTR, position, and freshness. - `get_business_profile`: Google Business Profile discovery, clicks, calls, directions, and keywords (beta). - `get_web_analytics`: Connected GA4 traffic, acquisition, pages, events, conversions, and setup. - `get_site_insights`: Saved site crawl or combined SEO, GEO, GSC, profile, and AI summary. - `get_page_insights`: List or read saved page analyses, with full content as a resource. - `manage_page_scans`: Delete one or many saved page scans by analysis ID. - `search_sitemap`: Search your site or a competitor sitemap by topic. - `get_robots_txt`: Compact crawler rules for search and AI user agents. - `get_domain_rating`: Current Ahrefs Domain Rating with required attribution and license link. - `get_page_speed`: Live Google PageSpeed Insights and Lighthouse scores, CWV, and diagnostics. - `find_opportunities`: Highest-priority actions across search, crawls, AI, and competitors. - `run_analysis`: Start a site, page, visibility, AI presence, or competitor analysis. ### AI presence and competitors Measured AI answers, observed searches, Google fan-out, and competitor evidence. - `get_ai_presence`: Overview, prompts, runs, status, or one model result through a view. - `get_ai_answer`: Read one tracked model answer with mentions, citations, and search trace metadata. - `get_ai_search_trace`: Observed underlying searches from tracked ChatGPT and other model results, when exposed. - `get_ai_search_opportunities`: Google AI Overview and SERP fan-out from saved PAA, related searches, and citations. - `get_community_sources`: Wikipedia, Reddit, Quora, YouTube, Facebook, LinkedIn, and review pages cited in AI answers (overview, urls, lookup). - `get_competitor_insights`: Competitor list, scan, visibility gap, or content coverage through a view. - `manage_prompts`: Create, edit, archive, or restore prompt groups and buyer prompts. - `manage_competitors`: Add, update, archive, or restore tracked competitors. ### Action Center Kanban The shared board used by your team, Actions chat, and MCP. Progress existing tasks first, not a backlog of AI notes. Use concise plain text, without Markdown. - `get_action_center`: Read the shared Kanban for agreed work. Find existing outcomes before creating more tasks. - `get_action_task`: Read one current task, its version, brief, next step, steps, and output. - `manage_action_center`: Create, improve, move, or delete one agreed outcome. Concise plain text only. Done tasks expire after 7 days. ### Legacy Action Center (read-only) For retrieving old work only, available until October 20, 2026. Use the new Kanban for all ongoing work. Do not copy the old backlog wholesale. - `get_work_items`: Temporary read-only Legacy Action Center access for saving or migrating chosen work before October 20, 2026. ### CMS and external actions Work with connected CMS content or request Google indexing separately from your task board. - `request_gsc_indexing`: Ask Google to crawl a URL on your property. - `get_cms`: Read optional Bloomiro CMS status, documents, page content, and schema. - `manage_cms`: Create, save, publish, unpublish, or delete Headless CMS pages when connected. ## FAQ ### What is Bloomiro MCP? Model Context Protocol (MCP) is an open standard for connecting AI assistants to tools and data. Bloomiro MCP is a server at https://api.bloomiro.com/mcp that brings your live Bloomiro project into Cursor, Claude, ChatGPT, Codex, and VS Code. Your AI can read the same Google and AI visibility data, scans, competitor gaps, and prioritized tasks that power Bloomiro's proprietary Optimization and Action Layer in the dashboard. ### How does MCP relate to Bloomiro's Optimization and Action Layer? The Optimization and Action Layer is Bloomiro's proprietary system on top of visibility analytics. The Optimization Layer is where teams view what to optimize and change. The Action Layer turns recommendations into workflows, next steps, and create actions. MCP does not replace that layer. It connects your AI tools to the same project context so you can optimize and act inside Cursor, Claude, or ChatGPT. ### What data does Bloomiro MCP connect? Google Search Console (queries, pages, clicks, impressions), Site Scan and Page Scan findings, competitor scans, AI presence results across ChatGPT, Gemini, Perplexity, and Google AI Overviews, community citation sources (Wikipedia, Reddit, Quora, YouTube, Facebook, LinkedIn, reviews), prioritized SEO and GEO tasks, and sitemap search across your site or competitors. ### Why use MCP instead of copying exports? Spreadsheets and screenshots go stale the moment you paste them. MCP returns current project state: the task list after yesterday's deploy, GSC after the weekend sync, or a competitor scan from this morning. Your assistant cites real scores and URLs from your property instead of guessing. ### Who is Bloomiro MCP for? Founders shipping fixes in Cursor or Claude Code, SaaS marketers who research in ChatGPT but need live visibility data, agencies where each teammate connects their own MCP key to shared org projects, and teams doing GEO who want AI mention checks next to traditional SEO tasks. ### How does MCP handle access and privacy? MCP connects to your Bloomiro account via OAuth or your own API key. Action Center tasks are shared with your organization, not public on the internet. Private tasks are not available yet. Other project data follows its access and visibility settings. You choose which site to authorize during connect. Multi-site users call get_project_context to resolve a domain or site_id, then pass site_id explicitly to later calls. ### How should an AI use the new Action Center? It is a shared Kanban for agreed work, not a place to dump research or chat transcripts. Read the board with get_action_center, inspect a task with get_action_task, and use manage_action_center to improve or move it. Keep titles, briefs, next steps, and outcomes concise, in plain text without Markdown. Update existing work before creating a distinct new task. Completed tasks stay in Done for 7 days before deletion. Save any output you want to keep. ### What do the Action Center statuses mean? To do is agreed work that has not started. In progress means someone is actively working on it. Review means an output is ready for someone to check. Done means the work is complete, with the outcome recorded. Move tasks as the work changes, not simply because the AI has read them. ### What happens to my old Action Center tasks? Legacy Action Center is read-only and available until October 20, 2026. Use get_work_items only to retrieve old work, not as the active board. In the new Action Center, open the three-dot menu beside Add and choose Open Legacy Action Center. Save what you need and move only work you still intend to act on into the new Kanban. If your MCP client does not show the new tools, refresh its tool list or reconnect. ### Which tools support Bloomiro MCP? Cursor, Claude (custom connector), ChatGPT (custom app with Developer mode), Claude Code, Codex, VS Code, and Perplexity connectors. OAuth is the default path. Some clients also support a local stdio server with a token from the connect page. ### Does MCP use credits? Most saved-data reads, including project context, tasks, GSC, crawler policy, and the compact MCP guide, are free. Fresh site, page, competitor, visibility, and AI presence analyses use credits under the same rules as the dashboard. run_analysis returns credits_used with the job result. ### How do I connect Bloomiro MCP? In your Bloomiro account, open your project's MCP tab in the dashboard. There is a setup guide for each supported client. Use https://api.bloomiro.com/mcp as the MCP server URL when your client asks for it. ### Can I connect ChatGPT or Claude to a headless CMS with Bloomiro? Yes. Bloomiro MCP can work with Bloomiro's Headless CMS so assistants in ChatGPT, Claude, or Cursor can draft, revise, and publish site pages with your live visibility context. For the CMS-focused walkthrough, see the guide on connecting ChatGPT and Claude to Bloomiro CMS. MCP itself connects your SEO and AI visibility project; CMS publish is a separate capability when Headless is connected. ### Is Bloomiro MCP included in pricing? Yes. MCP access is included on Builder and Growth plans at no extra fee. You still need a Bloomiro subscription and project; MCP connects to your own account, not a shared demo workspace. ### What can I ask after connecting? Examples: "What are my open SEO tasks?", "How do we compare to competitors on AI visibility?", "Which Reddit or YouTube pages do AI answers cite for us?", "Scan this URL for GEO issues", "What queries drove impressions last month in GSC?", or "Run AI presence on our buyer prompts." ### How do I read Community Sources from MCP? Ask for a Community Sources overview first, then the page list for a platform when you need URLs. Prompt text is optional so responses stay small. It reads your latest completed AI presence run and does not start a new crawl by itself. ## Links - Use in your AI: https://bloomiro.garden/agents - Pricing: https://bloomiro.garden/pricing - FAQ: https://bloomiro.garden/faq