Customizable, collaborative no-code, low effort AI platform for building, branching, and merging multi MCP agent & its workflows. Unlike most tools made for engineers, this is designed for creators who vibe with code.
Real-time Collaboration
Work together to build, maintain, test agents in real time.
No-Code Workflow Builder
Visually create AI logic using drag-and-drop blocks without needing to write code.
Customizable Agent Logic
Choose how your agents think and act. Combine web tools, rules, memory, file browsing, merge, branch and more.
Extensible Tooling & APIs
Connect your agents to external APIs and tools enabling them to take real-world actions programmatically.
System Info
v1.0.1
Hey everyone, Thank you for checking out this early version of the project. Right now, due to hardware constraints (just a single RTX 3070 8GB), I'm only able to support lightweight models like Gemma3:12B. Below is the roadmap that shows what’s next and where we’re headed.
Roadmap
0. Infrastructure
Seeking financial support to scale infrastructure.
1. More Model Support
- General: Qwen, Dolphin-Mixtral, llama, etc.
- Image generation: WAN-VACE, IC-Edit, Stable Diffusion 3
- Video generation: Sora, Veo, AnimateDiff, LTX-Video, WAN 2.1
2. B2B Integration
Extensible architecture for embedding into other apps and platforms.
3. Multimodal: TTS ↔ STT
Add support for voice chat via speech-to-text and text-to-speech.
4. Block Optimization
Improve the efficiency and speed of each block in the workflow.
5. Deep Thinking Options
Add toggles like “Deep Research” and “Think Longer” to guide agent depth.
6. Real-time Collaboration Improvements
Enhance live editing, chat syncing, and typing presence features.
7. Custom Blocks
Let users create, save, reuse, and share their own custom blocks (including agents as blocks).