Hey HN,

We are Phinite. We build infrastructure for deploying production-ready multi-agent systems.

The problem: building multi-agent systems requires you to think in graphs — nodes, edges, agent roles, tool connections, prompts. Most engineers spend more time architecting the workflow than actually solving the business problem.

So we built Phinite Aura — you describe what you want in plain English, and it designs the full multi-agent workflow for you: agents, prompts, tool connections, and the underlying code.

Example from our own testing: "I want to build an e-commerce product comparison assistant"

Aura generated: an orchestrator agent, 3 parallel product data extractor agents, a comparative analysis agent, connected tools (scrapers, APIs), and all the prompts. In under 60 seconds.

You can then modify it conversationally — "add a third product" — and it updates the entire workflow graph including agent logic, variables, and connections.

The canvas is live, versioned, and deployable. Integrations include Salesforce, Slack, GitHub, Google Sheets, Notion, and 23 others out of the box. Built-in dev/UAT/prod environments with Kubernetes-isolated traffic.

We built this because we were frustrated watching engineering teams spend weeks on workflow architecture before writing a single line of business logic.

Happy to discuss the architecture, how Aura handles ambiguous instructions, or where it still fails.

Try it: www.phinite.ai

Woaah this is cool.. this is like an AI infra for agents

Woah this is cool - its like an AI infra for agents