Skip to content
ATAI Today Brief
HomeNewsConceptsGuidesToolbox
AboutSubscribeUA
Subscribe

AI Today Brief

The daily AI-engineering brief. Built in public. EN · UA.

XTelegramLinkedInYouTubeRSS
NewsConceptsGuidesSubscribeAdvertiseAboutEditorial policyAI disclosurePrivacyTerms

© 2026 AI Today Brief. All rights reserved.

  1. Home/
  2. News/
  3. Vibe coding workflow/
  4. Grepathy Auto-Documents Claude Code Architecture Decisions directly in Git
Vibe coding workflow

Grepathy Auto-Documents Claude Code Architecture Decisions directly in Git

July 15, 2026· 4 min read
OKCurated by Oleksandr Kuzmenko, AI Product Engineer·Updated July 15, 2026·Sources cited on every story
AI-assisted · editor-reviewed·How we use AI
Grepathy Auto-Documents Claude Code Architecture Decisions directly in Git

Grepathy parses local Claude Code transcripts via git hooks to extract undocumented agent decisions and saves them as Markdown files. This solves the 30-day retention limit of Claude transcripts, ensuring teammates and future agents understand the architecture.

Impact: High

Why it matters

You can now review and preserve every silent decision your coding agent makes without interrupting your vibe-coding workflow.

TL;DR

  • 01Claude Code deletes your chat history after 30 days, losing crucial decision context.
  • 02Grepathy automatically extracts these decisions upon git push.
  • 03Historical summaries are fed back to future agents via pre-tool hooks.

Key facts

Runtime Dependencies
0 (Pure TypeScript)
Transcript Retention
30 days (Claude Code default)
Supported Tools
Claude Code (Codex upcoming)

The Vanishing History Problem

Coding agents like Claude Code perform dozens of micro-decisions during a feature build. By default, Claude Code deletes local session transcripts after 30 days, leaving developers with zero context when reviewers ask why a certain architecture path was chosen.

How Grepathy Automates Context

Instead of asking the LLM to write logs mid-task, Grepathy hooks into git push. It parses local chat histories, filters secrets, and distills the logic into .ai/why/ directory. If an agent edits a file that matches a touched path glob, Grepathy's PreToolUse hook automatically injects the historical reasoning back into the active prompt context.

Evaluation Results and Integration

In blind evals, agents supplied with Grepathy's why-pack significantly outperformed baseline runs by avoiding hallucinations regarding undocumented design patterns. It requires zero servers and relies purely on TypeScript with zero runtime dependencies.

Try it in 2 minutes

npm install -g grepathy
grepathy init

bash

✓ When to use

  • When using Claude Code for continuous autonomous development
  • To share agentic design decisions with human code reviewers
  • To prevent long-term transcript data loss on local machines

✕ When NOT to use

  • If you only use web UI-based LLM chats
  • For non-git projects with no version control workflow

What to do today

  • →Install grepathy globally
  • →Run grepathy init inside your active Claude Code project
  • →Verify the pre-push hook configuration

What the community says

  • “Amazing to me how little these companies respect the historic value of the transcripts. Most of my work is captured in them now.”

    — simonw on Hacker News

#Claude Code#Git#TypeScript

Sources

  • Grepathy GitHub Repository
ShareShare on XShare on LinkedIn
← Previous storyShare Model Context Protocol Servers Across 22 Clients with Toolport GatewayNext story →How Domain-Specific Languages Restrict LLM Outputs to Guarantee Reliable Code Generation

Related stories

  • Vibe coding workflowShift from Code-Review to Idea-Centric Design in AI-Driven Workflows
  • Vibe coding workflowRoute Claude Code Subagents to GPT-5.6 Sol via Proxy
  • Vibe coding workflowEvaluating Twelve Frontier Models in a Multi-Run App Build-Off
  • Vibe coding workflowStop Letting AI Write Lazy, Duplicated Code Patterns in Your Repository

Email digest

Get the morning AI brief

One email a day — the stories that matter for engineers, founders and tech leads. Human-edited, with links to primary sources.

  • ✓120+ sources scanned daily
  • ✓Edited by a human
  • ✓1 email per day
  • ✓EN + UA

By subscribing you agree to the privacy policy.