Clipnote
What you can do over MCP公開
What you can do over MCPの本文を直接見る

# What you can do over MCP Once your AI client is connected to Clipnote ([Connecting an AI client to Clipnote](https://clipnote.paritto.dev/p/65e19251-b6d3-451c-a746-067fa27c5842)), your AI can do five things with it. There are no commands to memorize. **Ask in plain language and your AI picks the right operation itself.** What follows is what it can and can't do, with examples of how to ask. ## The five operations | Operation | What it does | | --- | --- | | Find clips | Searches and lists your clips by title. Doesn't read the content | | Read a clip | Reads one clip's content in full | | Save a clip | Creates a new clip | | Update a clip | Replaces an existing clip's content | | Upload an image | Adds an image to your image assets and returns its URL | ### Finding clips > "List the meeting notes I've saved in Clipnote." > "Do I have a clip with Kyoto in it?" What comes back is the title, format, visibility, and last-updated time — **not the content**. That's so a list of dozens of clips doesn't bury the conversation in text. Search matches on the title, as a partial match. The content itself isn't searched. ### Reading a clip > "Read my Kyoto itinerary clip and let's carry on from there." > "Use those meeting notes to draft a thank-you email." This loads the clip's content. Your AI is also told **which collections the clip belongs to**, with their names and descriptions — a collection's description often turns out to be the clue that gives the clip its context. Find and read usually happen back to back, so there's no need to ask for them in two steps. ### Saving a clip > "Save this to Clipnote." > "Save this as public and tell me the URL." Your AI states the content, the title, and the format (HTML, Markdown, or plain text) explicitly. Unlike the dashboard, **nothing is auto-detected**. If it lands in the wrong format, say so — "save that as Markdown." If visibility isn't specified, the clip is saved **private**. When you do ask for public, your AI hands you the public URL right there, so you can pass it on without leaving the conversation. **You can't assign a collection while saving.** Do that from the dashboard (see below). ### Updating a clip > "Update that clip with this corrected version." > "Just fix the date on that invitation page." This **replaces the content in full**. There's no such thing as a partial edit; your AI assembles the whole new text and sends it. Whatever was there before is set aside in the version history automatically, so an unwanted result is always reversible ([Version history and restoring a version](https://clipnote.paritto.dev/p/31f6542c-c7b3-4ce9-9cc8-86ae8f58b648)). History entries are badged "MCP (OAuth)" or "MCP (API key)", so you can tell later **where an update came from**. ### Uploading an image > "Upload this image to Clipnote and put it on that page." This adds the image to your image assets and returns its public URL. Your AI can then embed that URL in whatever clip it saves or updates next. An uploaded image can be opened by anyone who knows its URL, regardless of the clip's visibility — see "What you need to know" in [Using image assets](https://clipnote.paritto.dev/p/af717712-f644-40bb-b51c-19eb737dea9f). ## What it can't do Worth being explicit about where the line is. Everything below is done **from the dashboard**. | Not available over MCP | Why | | --- | --- | | Creating collections, or changing what's in them | So your AI doesn't guess where something belongs and quietly reorganize your library. It can **read** collection membership | | Deleting a clip, or restoring one from the Trash | So nothing weighty happens on conversational momentum | | Archiving and unarchiving | Same reason | | Changing only the title | An update is always a replacement of the content | | Switching visibility after the fact | It can be set when saving, but changing an existing clip is a dashboard job | ## Clips your AI can't see AI clients only ever work with **active clips**. - **Archived** clips: not listed, not readable, not writable - Clips in the **Trash**: same If there's a clip you don't want your AI touching but don't want to lose either, archiving it is the sure way ([Saving and organizing clips](https://clipnote.paritto.dev/p/2813a01c-5ba9-4bca-ae90-ef4f76d3cc85)). So their existence can't be inferred, these come back as "not found" rather than "not allowed." ## Limits The same limits as the dashboard apply. - Clip content: up to 1 MB - Images: up to 5 MB, PNG / JPEG / GIF / WebP Go over and your AI client gets an error stating the limit and the actual size. Usually it will offer to split or compress on the spot. ## If you're worried about mistakes Some clients **ask for confirmation before every write** — ChatGPT's developer mode, for instance. Even with confirmations turned off, content updates always land in the version history and deletion isn't possible at all, so nothing ends up truly unrecoverable. If a conversation accidentally produces a pile of clips, delete them from the dashboard. You have 30 days to change your mind ([The Trash and undoing a deletion](https://clipnote.paritto.dev/p/58ab1de8-9922-4db6-91a0-abc46bdeb96c)). ## Read next - [Connecting an AI client to Clipnote](https://clipnote.paritto.dev/p/65e19251-b6d3-451c-a746-067fa27c5842) - [Saving and organizing clips](https://clipnote.paritto.dev/p/2813a01c-5ba9-4bca-ae90-ef4f76d3cc85) - [Version history and restoring a version](https://clipnote.paritto.dev/p/31f6542c-c7b3-4ce9-9cc8-86ae8f58b648)