Agora

Inspiration

What it doesAgora is a workflow orchestration platform with built-in AI telemetry for debugging LLM applications. Inspired by production failures that print() debugging can't solve, it makes AI workflows observable by default.

Core Architecture Agora uses a three-phase node model (prep → exec → post) for composable workflows:

text ┌──────┐ ┌──────┐ ┌──────┐ │ PREP │ → │ EXEC │ → │ POST │ │ Read │ │ Core │ │ Write│ │state │ │ logic│ │+Route│ └──────┘ └──────┘ └──────┘ PREP: Extract/transform shared state data

EXEC: Pure business logic (LLM calls, compute)

POST: Store results, return routing action

Key Features Node chaining: flow.start(A()) >> B() >> C()

Conditional routing: A() - "approved" >> SuccessNode()

Shared state: Single dict accessible by all nodes

Auto-retries: Built into every node

Full telemetry: Spans for each phase + LLM metrics + business context

Local-first: JSONL export, zero-config setup

Real Example python flow.start(Validate()) - "ok" >> Process() - "fail" >> Reject()

Built With

• asyncio
• cerebras
• cytoscape.js
• fastapi
• opentelemetry
• postgresql
• python
• react
• supabase
• tailwindcss
• traceloop
• typescript
• vite