Open-source Ai-workflow cuts coding agent token waste via zero-grep rules and persistent knowledge base
No_Professional_4310 · reddit · 2026-08-26
The author released Ai-workflow, an open-source, agent-agnostic workflow designed to reduce token consumption for AI coding agents like Claude Code, Cursor, and Copilot.
Key Features:
- Unified Config: A single AGENTS.md file serves as the source of truth for all agents.
- Zero-grep Rule: Agents must query pre-computed indexes (symbols, endpoints) before falling back to rg searches.
- Laned Routing: Tasks are routed into answer, small, or full lanes based on complexity. The full lane follows a Plan → Build → Review process with a 30-line cap on handoff docs to limit bloat.
- Persistent Knowledge Base: Uses an Obsidian vault to track resolved incidents, preventing agents from repeating mistakes.
Currently implemented in PowerShell, with Bash support in development.
More from coding & agent
- Antonio Gulli open-sources 'Agentic Design Patterns' guide with 21 chapters of code — tom_doerr · 2026-08-26
- Tutorial: Building a Voice-Powered Insurance Claims Agent with Twilio, OpenAI, and Box — craigsdennis · 2026-08-26
- AgentFlow lets small models outperform large ones via specialized roles — TheZachMueller · 2026-08-26
- Devin powers Rohlik Group's $1.3B grocery delivery — premqnair · 2026-08-26
- How Poke stores agent memories in Git using Code.Storage — marvinvonhagen · 2026-08-26
- Sentinel tool detects MCP server security flaws via static analysis — OpenAIDevs · 2026-08-26