AgentFramer

agentframer.com

AgentFramer is an agentic, MCP-based platform that lets AI agents generate images, videos, and audio files.

Connect your AI assistant to a suite of media generation tools through the Model Context Protocol. Describe what you want — the agent handles the rest.

What it does

• Image generation — generate images from prompts via connected AI models
• Video generation — create short-form and long-form video from text or image inputs
• Audio generation — produce music, voice, and sound effects through AI audio models
• MCP-native — works with any MCP-compatible AI assistant (Claude, Cursor, Windsurf, etc.)
• Agentic workflows — chain generation tasks together in multi-step agent pipelines

What is in this repo

• TanStack Start app for AgentFramer
• auth, billing, onboarding, workspace, and dashboard flows
• MCP server and CLI
• media generation integrations (image, video, audio)
• asset management and output storage
• generation history and workspace views

Hosted service

AgentFramer is available at agentframer.com. This is the only supported way to use AgentFramer. It includes:

• managed deployment and operations
• pre-configured AI provider credentials
• managed storage for generated assets

Stack

• React 19
• TanStack Start
• Cloudflare Workers
• InstantDB
• Better Auth
• Stripe
• MCP SDK

MCP CLI

The package exposes an agentframer CLI for MCP setup.

Initialize MCP config

npx -y agentframer@latest init

If you install AgentFramer directly from an MCP registry, you can usually skip manual API key setup. On first use, the server opens a browser and walks you through sign-in automatically.

For local development inside this repository:

node bin/agentframer.mjs init

There is a sanitized example config at .mcp.json.example.

Available MCP tools

• generate_image — generate an image from a text prompt
• generate_video — generate a video from a text or image input
• generate_audio — generate music, voice, or sound effects
• list_generations — list past generation jobs
• get_generation_status — check the status of an in-progress job
• get_generation_output — retrieve the output URL of a completed job
• submit_feedback

The tool schema is defined in src/lib/agent-mcp.ts.

Current limitations and notes

• video and audio generation jobs are asynchronous — poll with get_generation_status
• generated assets are stored in R2 and served via public URLs
• generation availability depends on the underlying provider's uptime and rate limits

Project policy

• License: MIT
• Security reporting: SECURITY.md