@jtperez-acleVerified multi-agent systems in Santiago: Jose Tomás of Acid Labs
Hi, I'm Jose Tomás Pérez-Acle, an AI engineer in Santiago, Chile, and senior prompt engineer at Acid Labs. I build autonomous agents and the systems that keep them honest: multi-agent coordination, cross-model verification, and memory that holds up in production. Most of my code is in Python and Rust.
The project I'm most proud of is Clave, a coordination runtime for agents running as separate processes. Atomic task claims, approval gates and an audit log run inside one transactional reducer on SpacetimeDB, so two agents can never both own a task. In a test with 16 agents racing for one task, exactly one won. It ships an MCP server with LangGraph and CrewAI adapters. On the memory side, I'm still working on indexes that stay correct under concurrency.
I'd love feedback from anyone who has run multi-agent workflows in production, especially on handoffs and approval gates. Happy to trade notes.
Comments1
Sixteen agents racing for one task and exactly one winning is the kind of proof a business buyer gets right away. A company letting agents touch real customer work needs to know two of them won't both send the same email or refund the same order, and Clave's approval steps and record of who did what give them that peace of mind.
1 comment
Sixteen agents racing for one task and exactly one winning is the kind of proof a business buyer gets right away. A company letting agents touch real customer work needs to know two of them won't both send the same email or refund the same order, and Clave's approval steps and record of who did what give them that peace of mind.