Pics.io MCP is a way to connect your AI assistant (Claude, ChatGPT, or Gemini) directly to your Pics.io library. Once connected, you can search for assets and view them right in your chat with the AI — no need to switch between tabs.
Instead of opening Pics.io, navigating to the right collection, and setting filters manually, you can simply ask your AI assistant:
💬 "Show me all PNG banners uploaded in the last 3 months"
The AI will find the matching assets in seconds and display them in the chat — with previews, names, and metadata.
What is MCP?
MCP (Model Context Protocol) is an open standard that allows AI assistants to connect to external systems and work with real data. Pics.io provides a remote MCP server so that your AI tool can work directly with your media library — there is no local server for you to run or maintain.
MCP is not a new product or a plugin — it is a bridge between your AI assistant and Pics.io.
What Pics.io MCP can do (Version 1)
With Pics.io MCP connected, your AI assistant can:
Get a list of assets from your library
Search for assets by text query, file type, or collection
List and search your Pics.io collections by name — a good first step in large libraries, to narrow down where relevant assets are located before searching assets directly
Search in natural language via a prompt (e.g. "find all videos with cats")
View full metadata of a specific asset
View asset previews right in the chat
Version 1 includes four tools, all of them read-only:
Tool | What it does | What it returns |
search_assets | Finds assets by a text query (natural-language description, file name, keywords), with optional filters: file type, collection, and which fields to search in | Asset ID, name, file type, preview, collection, basic metadata, and a link back to Pics.io |
search_collections | Lists your Pics.io collections, or searches them by name (and by description, where available), so the AI can identify the right part of your library before searching individual assets | Collection ID, name, description, parent collection/hierarchy (where applicable), and asset count where available |
get_asset | Retrieves full information about a specific asset | Full metadata, keywords, description, AI-generated tags (if any), collection path, dimensions/duration, and a link |
get_asset_preview | Fetches a lightweight, safe preview of the asset so the AI can "look" at it | A preview image embedded right in the chat or a fallback preview link, depending on the AI provider |
Using collections to navigate large libraries
Collections in Pics.io usually represent the main way assets are grouped — by project, campaign, client, or event — and often carry names and descriptions that describe what's inside. Your AI assistant can use the search_collections tool to check these collections first, before searching assets, which helps it narrow down where to look in large or highly structured libraries.
For example, if a collection named "Spring 2026 Campaign" has the description "Final approved lifestyle photography for the Spring 2026 European marketing campaign," a request like:
💬 "Find all approved images from our Spring European campaign"
can be resolved by first identifying that collection, then searching assets within it — instead of searching the entire library.
As with all Pics.io MCP tools, collection discovery respects your existing Pics.io permissions: the AI can only see collections your account already has access to.
What Pics.io MCP cannot do (Version 1)
Pics.io MCP works in read-only mode. Your AI assistant cannot:
Upload files
Change metadata or keywords
Delete or move assets
Create sharing links
Change access permissions
Access original files — only generated previews are shared, never the originals
⚠️ Version 1 works in view-only mode. No changes can be made to your library through MCP.
Supported AI tools
Pics.io MCP works with Claude, ChatGPT, and Gemini. You can add Pics.io as a custom connector in any of the three — the connection details (server URL, sign-in) are the same.
How to connect Pics.io MCP
Pics.io MCP can be added the same way in Claude, ChatGPT, or Gemini: as a custom connector, using the same connector details in all three — Name: Pics.io, Server URL: https://mcp.pics.io/mcp, no manual Client ID/Secret needed.
Only the menu you use to add a custom connector differs between the three apps, and those menus can change over time — if a step below doesn't match what you see, check that assistant's own help documentation for the current process.
Claude Desktop
What you'll need:
Claude Desktop installed on your computer
Access to log in to your Pics.io account in a browser
In Claude Desktop, open Settings → Connectors (on some versions this is under Customize → Connectors).
Click Add custom connector.
Enter a name, for example Pics.io, and the server URL:
https://mcp.pics.io/mcp. Leave the Advanced fields (OAuth Client ID / Secret) empty — Pics.io MCP uses Dynamic Client Registration.Click Add, then click Connect.
A browser window opens — log in to your Pics.io account if needed, then approve the consent screen.
Fully restart Claude Desktop (quitting the process from Task Manager/Activity Monitor is more reliable than just closing the window) and open a new chat.
A few things to keep in mind:
On the Claude Free plan, only one custom connector can be added at a time.
For Claude Team/Enterprise accounts: a workspace admin can add the Pics.io connector at the organization level. Once that's done, other members can simply click Connect without repeating the setup steps above.
NB! If you're stuck in a login loop after clicking Connect, make sure you have an active Pics.io session in the browser where the OAuth redirect opens — log in there first, then try Connect again.
ChatGPT
Custom connectors in ChatGPT require a paid plan (Plus, Pro, Business, Enterprise, or Edu) and are added from ChatGPT's own connectors menu.
In ChatGPT, open Settings → Apps & Connectors.
If you haven't already, go to Advanced and turn on Developer mode (a one-time step per account).
Back in Apps & Connectors, click Create (may also appear as Add custom connector).
Enter a name, for example Pics.io, and the server URL:
https://mcp.pics.io/mcp. Choose OAuth as the authentication method, then save/create the connector.Start a new chat, open the connectors/tools menu, and turn the Pics.io connector on.
When prompted, log in to your Pics.io account and approve the consent screen.
Gemini
How you connect Pics.io in Gemini depends on whether you use the Gemini web/mobile app or the Gemini CLI (terminal).
Gemini web app (Spark)
Adding a custom MCP connector in the Gemini app requires a personal Google account (not a Workspace/work-school account) on a Google AI Pro or Google AI Ultra plan — it isn't available on the Gemini Free plan. Availability is still rolling out by region: Pro-tier access is US-only for now, and Ultra-tier access is available in most countries except the EEA, UK, Switzerland, and Nigeria.
In the Gemini app, open Settings & help → Connected Apps (or go to gemini.google.com/apps).
Switch to the Other tab and scroll to the bottom.
Select Add a custom app.
Enter the server URL:
https://mcp.pics.io/mcp, then click Next.Gemini opens a Pics.io sign-in window — log in and approve access.
Turn the Pics.io app on and start a chat.
Gemini CLI
Google retired the individual-tier Gemini CLI on June 18, 2026 in favor of the new Antigravity CLI; only accounts on a Gemini Code Assist Standard/Enterprise license (or using API-key auth) keep access to the classic CLI. Pics.io publishes an MCP extension for the Gemini CLI ecosystem.
Make sure you have Gemini CLI (Code Assist Standard/Enterprise) or the Antigravity CLI installed.
Run command:
gemini extensions install https://github.com/TopTechPhoto/picsio-mcp-plugin
Follow the sign-in prompt to connect your Pics.io account.
Gemini Enterprise (corporate accounts)
Organizations on Gemini Enterprise connect Pics.io differently: a workspace admin adds it as a custom MCP server from the Gemini Enterprise admin console (Manage team → Connected apps → Add MCP Server), entering the Pics.io server URL. This is a separate flow from the personal Gemini app above and requires admin access.
Technical Limitations
Search and metadata (search_assets, get_asset):
No restrictions on asset type or size — metadata is returned for any asset in your library
Results are paginated: 1–100 assets per request (20 by default)
Requests time out after 30 seconds
Collections (search_collections):
Only collections your Pics.io account has access to are returned — the same permission rules that apply in the Pics.io interface apply here
Previews (get_asset_preview):
Supported formats: images, videos (a preview frame is used), and PDFs. Other formats (archives, Office documents, audio, etc.) cannot be previewed — the AI will receive an explanatory message instead
Size: the 25 MB limit applies to the generated preview thumbnail (JPG), not the original file. A 300 MB video or a large RAW/HEIC file will still return a preview without issues — only the small JPG is embedded in the chat. In the rare case the thumbnail itself exceeds 25 MB, the AI receives a temporary preview link instead
Originals are never exposed — only the generated preview is shared with the AI
Example Prompts
Here are a few ways teams use Pics.io MCP in everyday work:
Marketing: "Show me all PNG banners uploaded for the spring campaign"
Design: "Find all photos of SKU PRD-2024-007 and show me their previews"
Content review: "Show me all assets from the 'Product Launch May 2026' collection"
Agencies: "Are there any horizontal banners tagged 'digital' in the Brand X collection?"
Large libraries: "Which collection would have our approved Spring European campaign photography? Search within it for anything portrait-orientation."
FAQ
Can the AI delete or modify my files?
No. Pics.io MCP works in read-only mode. The AI can only search for and view assets — no changes are made to your library.
Are my files transferred to the AI?
MCP passes asset metadata and generated previews for display in the chat. Original files are never exposed — only lightweight previews. Access is authorized via OAuth, so the AI only sees what your Pics.io account has access to.
Do I need to install anything?
For Claude, ChatGPT, and the Gemini web app, no — you just log in to Pics.io through your browser when prompted, with nothing to install. The one exception is Gemini CLI, which requires running gemini extensions install https://github.com/TopTechPhoto/picsio-mcp-plugin to add the Pics.io extension.
Can the AI see files I don't have access to in Pics.io?
No. MCP uses your Pics.io login (OAuth) and respects all configured access permissions. The AI sees exactly what your account sees in the interface.
Can the AI browse my collections?
Yes. The AI can list your Pics.io collections or search them by name (and description, where available) to figure out where relevant assets are likely located, before searching individual assets. As with everything else in Pics.io MCP, it only sees the collections your account has access to.
How do I disconnect or revoke access?
Remove the Pics.io connector from that assistant's own settings — for Claude Desktop, Settings → Connectors; for ChatGPT, Settings → Apps & Connectors; for Gemini, Connected Apps.
Is MCP activity recorded anywhere?
Individual searches and asset views through MCP aren't logged in the Audit Trail. What is tracked is the connection itself — when a connector is set up — not which files were viewed afterward.
Which file types can the AI preview?
Images, videos (as a preview frame), and PDFs. Other file types (archives, Office documents, audio, etc.) can still be found via search and their metadata viewed, but a visual preview is not available for them.
Does using MCP affect my AI credits or limits in Pics.io?
No. Pics.io MCP uses standard API requests to your library — with no additional load on the platform's AI resources.
What is MCP, exactly?
MCP (Model Context Protocol) is an open standard that allows AI assistants to connect to external systems and work with real data. Pics.io provides a remote MCP server so that your AI tool can work directly with your media library.
If you have any questions about Pics.io MCP, feel free to reach out to us at support@pics.io or via the live chat.