Model Context Protocol
Grigora speaks the Model Context Protocol, so an agent can do the work in your site instead of describing it to you. It writes the post, builds the directory, fixes the redirects, and publishes when you say so. 42 tools, one connection, no copying and pasting between a chat window and a dashboard.
Every tool below is a real operation on your site, the same one the dashboard calls. Nothing here is a suggestion an agent hands back for you to carry out yourself.
Draft, edit, and publish posts. Content can be sent as HTML or as the same block tree the visual editor uses, so a post an agent writes stays editable by hand afterwards.
list_cms_postsget_cms_postcreate_cms_postupdate_cms_postpublish_cms_postunpublish_cms_postdelete_cms_postlist_cms_categoriescreate_cms_categoryupdate_cms_categorydelete_cms_categoryModel a collection, define its fields, then fill it. Fields can be text, rich text, image, link, datetime, switch, option, reference, map, HTML, multi-image, or video, so a directory holds real data rather than flat pages.
list_directory_collectionscreate_directory_collectionupdate_directory_collectiondelete_directory_collectionlist_directory_fieldscreate_directory_fieldupdate_directory_fielddelete_directory_fieldlist_directory_itemsget_directory_itemcreate_directory_itemupdate_directory_itemdelete_directory_itemRead the structure of a site, add pages, and update the ones already there. Every call is scoped to a project, so an agent works on one site at a time.
list_projectsget_project_detailslist_pagesget_pageadd_pageupdate_pageUpload images and files through signed URLs, then reference them from posts, listings, and pages.
list_assetsget_assetget_asset_upload_urlsdelete_assetsKeep URLs intact after a restructure, and read or replace the custom head and body code on a site.
list_url_redirectscreate_url_redirectupdate_url_redirectdelete_url_redirectget_custom_codeupdate_custom_codeRead traffic figures, and push everything live. Nothing an agent changes reaches visitors until a publish call runs, so drafting and shipping stay separate decisions.
get_analytics_umamipublish_projectAsk an agent to research a topic, draft the post, attach a featured image, assign categories, and publish it. The draft arrives as editable blocks, not a locked export.
Create the collection, define the fields it needs, then add an item per row. What would be an afternoon of form filling becomes one instruction.
Move content across, then create the URL redirects that keep the old addresses working, so search engines follow the change instead of hitting dead ends.
Point a scheduled agent at the server to refresh listings, publish queued posts, or update a changelog page, without anyone opening the dashboard.
Connecting
Your endpoint and key are specific to your workspace, so you copy them from the dashboard rather than from this page.
Open Settings in your Grigora dashboard and copy your MCP endpoint and key.
Paste them into your MCP client configuration, then restart the client so it picks the server up.
Ask the agent to list your projects, then work on the one you want. Every tool is scoped to a project id.
Asking for a directory of local venues turns into a sequence the agent runs on its own:
list_projectscreate_directory_collectioncreate_directory_fieldcreate_directory_itempublish_projectThe last call is the only one visitors ever see the result of. Everything before it stays a draft until you publish.
It is a Model Context Protocol server that exposes your Grigora site to an AI agent as a set of tools. Instead of describing what to change and copying the result into the dashboard by hand, the agent calls the same operations the dashboard calls: creating posts, building directory collections, adding pages, managing redirects, and publishing.
Any client that speaks the Model Context Protocol, which is an open standard rather than a Grigora format. That includes Claude Code, the Claude desktop app, and other MCP capable editors and agent runtimes.
Only if you ask it to. Creating or updating content and publishing a project are separate tools, so an agent can prepare everything and stop, leaving the publish call to you. Most MCP clients also ask for approval before running a tool that writes.
Yes. Posts can be sent as the same block structure the visual editor uses, so anything an agent creates opens normally in the editor and can be changed by hand afterwards. Nothing arrives as a locked import.
No. It is part of your Grigora account and works against the same site, content, and limits as the dashboard.
Build the site, connect the server, and let the agent do the part you were doing by hand.