How Greptile reviews every PR with sub-agents on Together AI: 50% faster, 2x cheaper
togethercompute · x · 2026-10-22
Greptile, an AI code review tool (YC W24, Series A led by Benchmark), shares its production architecture on Together AI: as more code gets written by AI, the first review pass must be independent of the tool that wrote it — and fast.
- The Greptile V5 agent spins up sub-agents on every pull request, pairing open and frontier models so one model's blind spots don't decide what gets caught
- It picks the best model per review task and works across every codegen platform including Cursor, Codex, and Claude
- Running open models on Together AI Dedicated Inference: 50% lower time to first token, 2x lower review cost
- Agents build semantic understanding of the whole repo so reviews reflect patterns, architecture, and intent
Related event: Greptile v5 Launches Multi-Agent Code Review for Every PR(2 posts)→
More from coding & agent
- "My job is telling Claude to think bigger"—devs joke about AI ambition gap — paulnovosad · 2026-10-22
- Superset MCP Server Lets LLMs Query Apache Superset via REST API — modelcontextprotocol · 2026-10-22
- Crates MCP Connector Wraps crates.io REST API, Free and Auth-Free — modelcontextprotocol · 2026-10-22
- Ropedia's S-Agent 2.0 Tops Spatial Reasoning Benchmarks With Zero Training, Scoring 66.5 on MMSI — CyberRobooo · 2026-10-22
- Eden Deploys Vercel's Eve Agent to Cloudflare With One Command, Fully Open Source — ritakozlov · 2026-10-22
- Ben Bajarin: rambling research notes into AI for organized workshop analysis works well — BenBajarin · 2026-10-22