Sobmit Documentation
Sobmit is a visual flow builder for AI agents. Instead of writing code, you drag nodes onto a canvas, wire them together with edges, and press Run. The result is a durable workflow that can call language models, search the web, call MCP tools, make HTTP requests, and generate images, video, audio, 3D models, tables, HTML pages and presentations.
What you can build
- Agents — prompt a language model, optionally with structured (JSON-schema) output.
- Research pipelines — a Web Search or Deep Research node that returns a cited report.
- Media pipelines — generate an image, turn it into a video, merge clips.
- Automations — trigger a flow from a webhook, run it on a schedule, or call it from an external MCP client.
How to read these docs
| You want to… | Start here |
|---|---|
| Build and run your first flow | Quick start |
| Understand nodes, edges, handles and runs | Core concepts |
| See complete flows you can copy | Use cases |
| Learn what a specific node does (with examples) | Node reference |
| Connect external MCP servers | MCP servers |
| Trigger a flow over HTTP | Webhooks |
| Run a flow on a recurring schedule | Schedules |
| Edit a flow with other people live | Realtime collaboration |
The node catalog at a glance
Every node belongs to one of three groups: Input, Tools and Output.
- Input —
Start,Prompt,Upload File,Agent,Router. - Tools —
HTTP Request,MCP,Web Search,Deep Research,Image Generation,Video Generation,Merge Videos,3D Generation,Audio Generation,Transcribe Audio,Table Generation,HTML Page,Presentation,Email,For-each,Approval,Iterate. - Output —
Output.
Each node has a dedicated page under the Node reference with its inputs, configuration and a concrete example.
Four shapes that cover most flows
Rather than starting from an empty canvas, start from a shape that already works. Use cases walks through all four — each ships as a built-in template and as a public flow you can open and inspect.
- Fan-out — one research pass produces a summary, a table and a deck.
- Routing — an agent returns a score, a condition on the edge decides what runs next.
- Upload → artifacts — a recording becomes a transcript, a summary and an action table.
- Schedule — a weekly check that reports even when nothing changed.