liminary

Name: liminary
Author: liminary

Your AI's memory for what you actually know: recall across your documents, notes, and meetings

От сообществаРаботаетБез входаГлобальныйБесплатноТолько чтение

Что умеет

Health: Check that the MCP connection and authentication are working end-to-end. Returns the authenticated Liminary user's id and the OAuth scopes granted to this client, and performs no library reads
Get Collections: Retrieve one or more collections by their ids, returning each collection's title, description, goal, and the list of document ids it contains. Find collection ids with search_library,
Create Collections: Create one or more collections in the user's Liminary library, each with a title and an optional description and goal (what the user wants to get out of the collection). Returns th

Какие данные видит

Нужен ли аккаунт

Не нужен: сервер работает без входа

Your AI's memory for what you actually know: recall across your documents, notes, and meetings

Список инструментов сервера (12)

Технические названия из tools/list. Нужны только разработчикам.

health	Check that the MCP connection and authentication are working end-to-end. Returns the authenticated Liminary user's id and the OAuth scopes granted to this client, and performs no library reads or writes. Call this first to confirm a working connection before using the other tools. Requires the `mcp:read` scope.
get_collections	Retrieve one or more collections by their ids, returning each collection's title, description, goal, and the list of document ids it contains. Find collection ids with search_library, then use get_documents to read members, add_to_collection or save_urls to add documents, and create_collections to make new ones. Unknown or inaccessible ids are returned as not found. Requires the `mcp:read` scope.
create_collections	Create one or more collections in the user's Liminary library, each with a title and an optional description and goal (what the user wants to get out of the collection). Returns the id of each created collection. Afterwards use add_to_collection or save_urls to populate a collection with documents. Requires the `mcp:write` scope.
add_to_collection	Add one or more existing documents to a collection by id. Documents already in the collection are skipped, and the result reports how many were newly added versus already present. The target collection must grant document-add permission; get document ids from search_library or get_documents and collection ids from get_collections. Requires the `mcp:write` scope.
get_documents	Retrieve the full content of one or more documents by their ids, including title, summary, key takeaways, source URL, and the complete markdown body. Use the fields argument to choose what to return (defaults to title, summary, and URL; request the body field for the full text). Get document ids from search_library results. Unknown or inaccessible ids are returned as not found. Requires the `mcp:read` scope.
save_urls	Save one or more web URLs to the user's library as documents. Each URL is fetched, scraped, and indexed so it becomes searchable within seconds, and can optionally be added to a collection in the same call. Returns per-URL success or failure. Use this to capture articles, papers, or web pages; to store free-form text you wrote yourself, use write_document instead. Subject to the user's document quota. Requires the `mcp:write` scope.
write_document	Create a new document in the user's library from finished markdown content you provide — supply the written text, not a prompt, as this tool does not generate content itself. Give it a title and the markdown body, and optionally a collection to file it under; it is saved as an artifact and indexed for search. Revise it later with edit_document, or use save_urls to capture web pages instead. Subject to the user's document and daily-upload quotas. Requires the `mcp:write` scope.
edit_document	Edit an existing artifact document in place. Provide exactly one of `edits` (a list of search/replace operations, where each search string must match the current body exactly once) or `new_content` (a full markdown replacement of the body), plus an optional new title. Only artifact documents created through Liminary can be edited; other document types are rejected. Read the current body with get_documents first so your search strings match. Requires the `mcp:write` scope.
get_notes	Retrieve one or more notes by their ids, returning each note's title and its content flattened into a single markdown body. Reads any note the caller can access, but structured notes such as highlights and image annotations lose their internal structure when read this way and are read-only via MCP. Set include_parent_context to also return the title of each note's parent document or collection. Find note ids with search_library; revise an editable note with edit_note. Requires the `mcp:read` scope.
save_notes	Create one or more notes attached to a parent document or collection. Each note has a title and markdown content and is saved as a simple note. Provide the parent id and parent type for each note; the parent must exist and grant note-add permission. Find parent ids with search_library, get_documents, or get_collections, and revise a note later with edit_note. Requires the `mcp:write` scope.
edit_note	Edit an existing note in place. Provide exactly one of `edits` (a list of search/replace operations) or `new_content` (a full markdown replacement of the body), and set return_new_content to get the updated body back in the response. Only simple single-body notes are editable; structured notes such as highlights, image annotations, and lookups are read-only and return a generic not-editable error. Read the current note with get_notes first so your search strings match. Requires the `mcp:write` scope.
search_library	Search the user's Liminary library by meaning, not just keywords. Accepts one or more natural-language queries and returns the most relevant documents, notes, and collections, each with excerpt snippets and a relevance score. Filter by item type, document type, a specific collection, or created/updated date ranges; omit the queries to list the most recently updated items instead. Follow up with get_documents, get_notes, or get_collections to fetch the full content of any result. Requires the `mcp:read` scope.