Pixelesq MCP¶
Remote MCP server (Streamable HTTP, OAuth 2.1 PKCE) - the official agent interface for Pixelesq websites. Agents manage a whole site from a conversation: create pages and build them section by section, write entries into collections, set SEO metadata and JSON-LD structured data, read Search Console performance and site analytics, manage redirects, forms, images and the theme. Every save lands in a draft; the live site changes only when the agent publishes, and a publish is bound to the revision it reviewed.
Server type: Remote (Streamable HTTP)
Auth: OAuth 2.1 with PKCE (Google sign-in, or Pixelesq email + 6-digit code)
Endpoint: https://mcp.pixelesq.app/mcp
Tools: 62 total (59 offered in Claude; 3 AI image tools client-gated)
Scopes: projects:read, content:read, content:write, publish, analytics:read
Pricing: Pixelesq plan (sign up at app.pixelesq.ai)
Category: Business Operations
Built by: Pixelesq (pixelesq.com)
Tools¶
| Tool group | Purpose |
|---|---|
| list_projects, get_project, list_domains, search_content | Projects, custom domains and their verification state, cross-site search over pages, entries and collections |
| list_pages, get_page, get_page_content, get_page_revisions, create_page, duplicate_page, update_page_meta, save_page_content, publish_page, unpublish_page, delete_page, revert_page_revision, revert_page_revision_to_version | Page CRUD with SEO metadata (title, description, Open Graph image, indexing preference, JSON-LD patched key by key) and revision rollback |
| list_section_types, get_section_defaults, get_section_content, save_section_content, insert_section, remove_section, reorder_sections, bind_section_to_collection, unbind_section_from_collection | Section-level page building from the section catalog; connect a section to a collection so it lists entries |
| list_collections, get_collection, list_entries, get_entry, create_collection, update_collection_schema, create_entry, save_entry_content, update_entry, duplicate_entry, publish_entry, delete_entry | Collections and entries: blog posts, team members, case studies, products. The agent reads the schema first so content matches the fields |
| list_partials, get_partial_content, save_partial_content, publish_partial, unpublish_partial | Header and footer partials, reusable across every page |
| get_gsc_connection_status, get_search_performance, get_page_index_status, get_all_page_statuses, inspect_url, get_site_analytics, get_ai_referral_analytics | Search Console performance and page index status, site analytics, and AI referral analytics |
| list_redirects, create_redirect, update_redirect, delete_redirect, list_forms, get_form_submissions | Redirect management, forms and their submissions |
| list_assets, search_stock_images, upload_asset, upload_image, get_theme, update_theme | Asset library, stock photo search and import, uploads from chat, and theme variables (colors, fonts, spacing, radius) as draft or published at once |
| generate_image, get_image_job, get_imagegen_quota | AI image generation (not offered in Claude, available in other clients) |
The draft-first model is the safety layer: a tool outside the granted scope answers with a reconnect request, saves are refused as stale when the page changed since the last read, and publishing requires explicit confirmation.
Connection¶
- Sign up at app.pixelesq.ai first - the connector needs an existing account.
- Claude: Settings, Connectors, Add custom connector, enter
https://mcp.pixelesq.app/mcp, sign in with Google or your Pixelesq email, then approve the permission grants. - Claude Code:
claude mcp add --transport http pixelesq https://mcp.pixelesq.app/mcp, then run/mcpto sign in. - Other clients (Cursor, VS Code, ChatGPT): add the server URL in the client's MCP config and complete the OAuth sign-in flow.
- Optional plugin:
/plugin marketplace add pixelesq/pixelesq-skills, then/plugin install pixelesq@pixelesq- adds seven skills and four specialist agents. - mcp.json form:
{"mcpServers": {"pixelesq": {"type": "http", "url": "https://mcp.pixelesq.app/mcp"}}}
Verification (Sep 12, 2026 morning sweep)¶
Endpoint live-probed over JSON-RPC initialize: HTTP 401 with OAuth protected-resource metadata and the exact scope string projects:read content:read content:write publish analytics:read - the auth gate the docs describe. The official directory listing carries a full 62-tool reference (grouped by projects, pages, sections, collections, partials, SEO/analytics, redirects/forms, images/theme) and best practices.
Integration with CorpusIQ¶
CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.
This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.