What is Neurovn?
Neurovn is an open-source visual canvas for designing multi-agent AI workflows and estimating their cost, token usage, and latency — before you write a single line of code or make a single API call.
Why Neurovn?
Visual Workflow Design
Drag-and-drop nodes onto an infinite canvas to compose multi-agent pipelines visually.
Instant Cost Estimation
Token, cost, and latency breakdown in under 10 ms — zero external API calls.
What-If Scaling
Project monthly costs at any production volume with sensitivity analysis.
Bottleneck Detection
Automatically flags nodes consuming disproportionate cost or latency.
Who is this for?
AI / ML Engineers
Design and cost-model multi-agent architectures before committing to code.
Product Managers
Understand the cost profile of proposed AI features without needing to read code.
Finance & Ops
Get monthly cost projections and per-run breakdowns for budget planning.
Architecture at a Glance
The frontend (Next.js + React Flow + Zustand) communicates with a lightweight FastAPI backend for estimation. The estimation engine is pure computation — native tokenization for token counting, a JSON pricing registry for 38 models across 7 providers, and graph analysis (Tarjan SCC, topological sort, critical path).
Frontend
Next.js 15 · React 19
React Flow · Zustand
Backend
FastAPI · Python 3.11+
Under 10ms execution
How Neurovn fits your delivery lifecycle
Neurovn supports planning before code, integration during implementation, and runtime verification after deployment.
Phase 1 · Design
Canvas Planning
Model architecture in the visual canvas before writing code. Validate shape, branch flow, and cost envelope early.
Phase 2 · Build
CLI + Decorator Integration
Import real workflows through CLI or instrument runtime functions with decorators to generate trace-linked canvases.
Phase 3 · Operate
Runtime Verification
Compare actual behavior against estimates, review drift, and refine model/provider choices with production evidence.
Choose a documentation track
Canvas Foundations
Learn graph concepts, node behavior, estimation logic, and scenario-driven workflow design.
Start with QuickstartIntegrations
Implement CLI and decorator paths, then apply framework-specific guides for OpenAI, LangGraph, LangChain, Google ADK, FastAPI, and CrewAI.
Open integration guides