Docs›Quickstart

Quickstart

Get your AI agent generating images, video, and audio in under five minutes using the Model Context Protocol.

01Setting up MCP

MCP (Model Context Protocol) is the open protocol that lets your AI assistant talk directly to AgentFramer's generation engine. Install the CLI and initialize the local bridge, that's all the wiring you need.

Terminal, zsh
$ npx agentframer@latest init
✓ Config created in .agentframer/config.json
✓ MCP local bridge started on port 4550

How jobs work

AgentFramer handles job queuing server-side, your agent calls a generation tool and gets back a job ID. It then polls for completion and retrieves the output URL when ready. No async infrastructure required on your end.

02Connecting your Agent

Once the bridge is running, register the AgentFramer MCP toolset with your AI client. You can connect via a locally running bridge for development, or point straight to our managed cloud infrastructure for production workloads.

Local Bridge

Connect to a locally running bridge for development and fast iteration cycles.

Managed Cloud

Scale to production with our hosted MCP infrastructure.

03Your First Generation

With the bridge connected, your AI assistant has access to generate_image, generate_video, and generate_audio. Just describe what you want, the agent handles model selection, queuing, and delivery.

›“Generate an image of a red sports car at sunset”
›“Which image, video, and audio models are available right now?”
›“Generate an image with nano-banana-2 at 1024 by 1024”
›“Upscale that image to 4 megapixels”
›“Generate a 5 second video of waves rolling onto a black sand beach”
›“Generate a 30 second ambient synth track for a futuristic product teaser”
›“Generate a square product image of a matte ceramic mug on white”
›“Show me my recent generations”
›“Open the last generated video”
›“How much balance do I have left?”
›“List my workspaces”
›“Create a workspace called Client Launch”
›“Switch to workspace wksp_123”
›“Invite alex@example.com to this workspace”
›“Show members and pending invites for this workspace”

Your AI assistant calls AgentFramer's tools on your behalf, no dashboard required. See the MCP overview for a full list of available tools and capabilities.

Was this page helpful?

How it works Inviting members