Adobe for creativity
Edit photos, cut video and design from templates via an AI agent.

618 MCP servers with 12,363+ tools, ready to connect to Claude, ChatGPT, Cursor and Agentman agents over the Model Context Protocol.
14 of 618 connectors in DesignShow all
Edit photos, cut video and design from templates via an AI agent.
Search, generate, import and export Canva designs.
Import media, edit it with the Underlord agent, and publish a shareable video link.
Pull design context, variables and Code Connect mappings into your agent, and make FigJam diagrams.
Publish real web apps from your AI assistant.
Render Mermaid diagrams inside a Claude conversation.
Search recording transcripts, cut and restyle edits, publish clips to seven social platforms.
Search Shutterstock's images, video, music and sound effects in any language.
Search Splice's royalty-free samples and download them from an AI agent.
Generate SketchUp .skp files from a conversation.
Search Unsplash photos, collections, illustrations and photographers.
Query streaming, social and chart data for millions of artists, tracks and festivals.
Create flowcharts, mind maps, wireframes and docs in your Whimsical workspace.
Search Taiwan's Central News Agency archive: 4.78M reports since 1990, an almanac since 2000, 3.37M photos since 1924.
Product names and logos belong to their owners. Each connector is built by its own developer, and Agentman is not affiliated with them. The ones Agentman builds and runs are marked “Built and run by Agentman”.
A server that exposes a tool’s capabilities over the Model Context Protocol, so an AI agent can call them directly. It is the standard way to let an assistant act in a system rather than just describe it.
Each server page carries its connection URL, which works in any MCP-compatible client. In Agentman Studio connecting is one click and there is no config file to edit.
Most are not. Each is built by its own developer, and each page names them and links to their documentation. The few Agentman builds and runs are marked “Built and run by Agentman”. Every page lists what the server exposes and cites the sources its factual claims come from.
Each page lists the tools the server exposes. Where we could read the server directly, every tool carries the safety label its vendor gave it: read-only, acts, or destructive. Each page also has a Limits section; read both before connecting anything to a production system.
Connect any of these once and your agents call their tools on their own: on a schedule, in a workflow, with nobody at the keyboard.