Releases·Tool Release
Dev adds 'why' command to ThoughtDAG to trace AI-agent conversations behind Git diffs in Claude Code, Codex repos
A developer added a read-only `why` command to ThoughtDAG, a local tool that recovers which AI-agent conversation produced a given code change. Running `npx thoughtdag why src/lib/api.ts` searches local agent transcripts for turns that changed or discussed the file and links back to the source turn. It addresses the gap where Git shows the diff but not the reasoning, rejected approaches, or assumptions behind agent edits.
4 engagement2 sourcesFri, Sep 4, 2026, 04:36 PM
AI Overview
- ThoughtDAG's read-only 'why' command searches local Claude Code, Codex, and canvas transcripts for turns that changed or discussed a file, linking back to source turns
- Recorded tool edits are marked Δ while explanation prose recovered from responses stays marked ≈ as candidate explanations
- The derived index stays local, source session files are never modified, and retrieved history is not automatically sent back to a model
Entities
Related
Releases·Tool Release
Show HN: Wy – Rust terminal tool surfaces agent context behind AI-generated code changes
A developer open-sourced Wy, a Rust terminal tool that surfaces agent session context behind AI-generated code changes, launching on GitHub and Hacker News. It aims to close the gap between what a git diff shows and the reasoning recorded during long-running agent execution.
A developer open-sourced Wy, a Rust terminal tool that surfaces agent session context behind AI-generated code changes, launching on GitHub and Hacker News. It aims to close the gap between what a git diff shows and the reasoning recorded during long-running agent execution.
1 engagement1 sourcehackernewsFri, Oct 9, 2026, 05:21 AM
Products·building
WTF — auto-review what your coding agent changed, locally
WTF gives developers an overview of what their coding agents actually changed, aiming to replace manual inspection of agent work. It runs fully locally with no API calls. The author is also experimenting with using it to ground agents in reality, though that is not working yet.
WTF gives developers an overview of what their coding agents actually changed, aiming to replace manual inspection of agent work. It runs fully locally with no API calls. The author is also experimenting with using it to ground agents in reality, though that is not working yet.
4 engagement1 sourcehackernewsSun, Sep 20, 2026, 10:05 AM
Products·building
CLI for Claude to record rationale behind git changes
diff-rationale is a CLI tool that integrates with Claude to record the reasoning behind code changes in git. It stages rationale records for each diff chunk, stores them in a separate git ref, and mints them into a journal when changes are merged, preserving the 'why' for future reference.
diff-rationale is a CLI tool that integrates with Claude to record the reasoning behind code changes in git. It stages rationale records for each diff chunk, stores them in a separate git ref, and mints them into a journal when changes are merged, preserving the 'why' for future reference.
7 engagement1 sourceredditFri, Aug 14, 2026, 03:12 AM
Products·launched
Grepathy: Distills AI coding agent decisions into local markdown
Grepathy is a tool that extracts decision-making transcripts from AI coding agents (like Claude Code) and saves them as markdown files committed with the code. It solves the problem of lost agent reasoning due to auto-deletion of transcripts, providing a local, serverless audit trail for developers who need to understand why their AI made certain changes.
Grepathy is a tool that extracts decision-making transcripts from AI coding agents (like Claude Code) and saves them as markdown files committed with the code. It solves the problem of lost agent reasoning due to auto-deletion of transcripts, providing a local, serverless audit trail for developers who need to understand why their AI made certain changes.
20 engagement1 sourcehackernewsWed, Jul 15, 2026, 01:26 PM
Releases·Tool Release
Show HN: ctx – Git blame that returns the original agent transcript
A developer released ctx (ctx.rs), a tool that extends git blame to trace any commit, file, line, or PR back to the exact agent tool call that wrote the code, then returns the full agent session transcript. The goal is to give coding agents fast local search over prior session transcripts so they can understand code provenance — why code was written, what tests were run — information too verbose for humans but useful for agents.
A developer released ctx (ctx.rs), a tool that extends git blame to trace any commit, file, line, or PR back to the exact agent tool call that wrote the code, then returns the full agent session transcript. The goal is to give coding agents fast local search over prior session transcripts so they can understand code provenance — why code was written, what tests were run — information too verbose for humans but useful for agents.
2 engagement1 sourcehackernewsWed, Sep 16, 2026, 02:46 PM