One change cut Claude Code tokens 3x: 10.4M→3.7M, $9.21→$2.81 with InsForge
Roger_M_Taylor · x · 2026-09-09
Developer @avichawla reports that adding InsForge Skills + CLI as a backend context-engineering layer for Claude Code cut token usage 3x.
Results:
- Before: 10.4M tokens, 10 errors, $9.21 cost
- After: 3.7M tokens, 0 errors, $2.81 cost
Why it works (based on Karpathy's context engineering principles):
- With Supabase, tables, RLS policies, auth providers, storage buckets, and edge functions each required separate calls; every response entered conversation history, bloating later context
- InsForge returns the full backend topology in one 500-token metadata call
- Instructions are split into narrow skills (CLI ops, debugging, SDK usage, integrations), loaded only as needed
- The CLI returns structured JSON with semantic exit codes for clear failure states
InsForge is an open-source all-in-one backend platform for agentic coding (12.9k GitHub stars).
More from coding & agent
- Frost agent autonomously runs cross-GPU benchmarks with onchain reasoning — Michael_Moroz_ · 2026-09-09
- Jitsu launches MCP server so agents can build, test and fix event pipelines — Aiden_Tech_Ai · 2026-09-09
- MCP tools can silently change schema after approval — an open-source proxy now catches it — jgarg27 · 2026-09-09
- KVMem pages agent context overflow to KV state, beating compaction on DeepSWE — omarsar0 · 2026-09-09
- Coding agents are the first product to compete with themselves — samgoodwin89 · 2026-09-09
- Local Qwen 27B on an RTX 3090 Builds a Playable Mario Clone From One Prompt — ChopSticksPlease · 2026-09-09