ThothFlow
Visual canvas to run multiple AI agents from a single prompt
6 followers
Visual canvas to run multiple AI agents from a single prompt
6 followers
Visual canvas. Type a task, AI agents do the work. No YAML. No config. Command Bar creates Workers across 13 providers from plain language. Parallel or chain. Secretary consolidates. Preflight Validator improves prompts and verifies accuracy. Test-first mode. SDD auto-generates specs. Local Agent Bridge: Tauri app gives agents sandboxed filesystem access via WebSocket. Safety: budget tracker, fallback, circuit breaker, encrypted keys. Audit log.
Hey Product Hunt! I'm Sérgio, co-founder of ThothFlow.
I built this because I wanted a more visual, dynamic, and automated way to work with AI agents. Instead of switching between chat tabs and configuring YAML pipelines, I wanted a canvas where I type a task and the agents do the rest.
How it works:
1. Bring your API keys
2. Open the canvas in your browser
3. Type what you want in plain language
4. The Orchestrator creates specialized Workers across 13 providers, they run in parallel or chain mode, and a Secretary consolidates the result
You can also connect the local agent bridge and let agents work directly on your files. Sandboxed, read/write access to your local folders from the canvas.
13 providers. One canvas. Zero YAML. Zero setup.
Wanna help shape the product? Test it, drop your honest feedback here in the comments, and I'll give you extended free access and a discount on the subscription. Just reach out on Discord (@sergiodjr) after commenting and I'll hook you up.
Happy to answer any questions!
I built this because I wanted a more visual, dynamic, and automated way to work with AI agents. Instead of switching between chat tabs and configuring YAML pipelines, I wanted a canvas where I type a task and the agents do the rest.
What makes ThothFlow different from other AI tools:
1. Command Bar — type a task in plain language, the Orchestrator auto-creates Workers with the right prompts across 13 providers. No manual config.
2. Preflight Validator — before execution, it validates and improves your prompts. Tool calls are shown inline so you see exactly what the agents are doing.
3. Secretary + Workers — Workers run in parallel or chain mode across different providers. A Secretary cross-checks and consolidates everything into a single result.
4. Local Agent Bridge — a desktop app (Tauri) that connects via WebSocket, giving agents sandboxed read/write access to your local files. Edit code, process CSVs, analyze documents directly from the canvas.
5. SDD + Test-first mode — generate a Software Design Document before implementation. Agents write tests first, then code. Production-ready from commit zero.
6. Safety stack — budget tracker, fallback orchestrator, circuit breaker, encrypted API keys. Five layers so your agents never surprise you with a bill.
7. Audit log — full history per interaction with tokens and cost. Blackboard Kanban shared between agents. Saved Agents library. Git panel. Agent monitor.
13 providers. One canvas. Zero YAML. Zero setup.
Wanna help shape the product? Test it, drop your honest feedback here in the comments, and I'll give you extended free access and a discount on the subscription. Just reach out on Discord (@sergiodjr) after commenting and I'll hook you up.
Happy to answer any questions!