Vesper
A visual platform for designing, running, and observing multi-agent AI workflows — node-based, multi-provider, with human-in-the-loop gates and end-to-end cost tracking.
Vesper is a full-stack AI workflow orchestration platform: a visual, node-based canvas for designing, executing, and monitoring multi-agent pipelines. It wires models from Anthropic, OpenAI, and Google into structured, observable workflows — a low-code automation studio for AI agents, with enterprise-grade observability and human oversight built in.
The canvas (React Flow plus a custom Zustand store with undo/redo and clean persistence) supports fourteen node types, from AI agents to control-flow primitives to approval gates. Execution runs on Inngest durable step functions, so long pipelines survive serverless timeouts, retry per step, run intra-level work in parallel, and can suspend on a human-approval gate for up to seven days before resuming.
Every run is fully observable: per-node tokens, latency, and cost to six decimals, a typed event log as the source of truth, cross-run analytics, and distributed LLM tracing via Langfuse. Users bring their own provider keys, encrypted at rest with AES-256-GCM and masked on read.
