iCreatAI News
Skip to content
Together AI Blog·· 2 days agoAI score58

How we make Together AI easy for coding agents to use

How we make Together AI easy for coding agents to use

AI summary

Together AI describes how it built an agent-evaluation system that runs common developer tasks against baseline, MCP, and skills configurations using multiple models, scoring runs with an LLM judge. The evals surfaced a mislabeled CLI sample and a skill file linking to a marketing page instead of the pricing table, plus an API error message that pointed agents to a nonexistent route; fixes took the GPU cluster task from 4 of 15 passes to 100%. The company also merged 14 separate skills into one unified skill, which initially dropped the pass rate to 75% before fixes brought it to 97.2%, and it plans to open-source its eval package next week.

Source: Together AI Blog · together.ai