bourbeau

6 ms·

1. 1.▲
   by bourbeau 6mo ago

   One pattern we have found helpful for agentic workflows: keeping the execution layer task-first (describe what you need, not which provider to call). You can layer runtime trust checks and cost constraints without the agent needing to know
2. 2.▲
   by bourbeau 6mo ago

   Interesting implementation. We have had better reliability by treating tool access as a capability-routing layer: request by task, not fixed provider, with explicit cost/latency constraints and runtime trust state (verified, reachable,
3. 3.▲
   My Customers are not Human. They're Agents. Here what they see(agoragentic.com)

   2 points by bourbeau 7mo ago|0 comments
4. 4.▲
   by bourbeau 7mo ago

   Thanks! We thought about escrow early on but went a different route — the platform holds the wallet balance and only settles after the seller's callback returns a valid response. If the seller times out, errors, or returns garbage, the
5. 5.▲
   by bourbeau 7mo ago

   I built a marketplace where AI agents discover and pay for each other's services via API. The demo shows what the full flow looks like from an agent's perspective — registration, browsing, invoking, and settlement. Everything runs
6. 6.▲
   Show HN: What an AI agent sees in an A2A marketplace – full API walkthrough(agoragentic.com)

   1 points by bourbeau 7mo ago|2 comments
7. 7.▲
   by bourbeau 7mo ago

   Hey HN, I built this. Quick ways to explore: Try the API right now: curl https://agoragentic.com/api/capabilities curl https://agoragentic.com/.well-known/agent-marketplace.json Agent-POV demo (ci
8. 8.▲
   Show HN: A marketplace where AI agents buy from other AI agents in USDC(agoragentic.com)

   1 points by bourbeau 7mo ago|1 comments
9. 9.▲
   Show HN: Agoragentic – Agent-to-Agent Marketplace for LangChain, CrewAI and MCP(github.com)

   1 points by bourbeau 7mo ago|0 comments