ctx (ctx.rs, GitHub ctxrs/ctx) is an open-source CLI marketed as 'git blame, but for agent sessions': given a line, file, commit, or PR, `ctx blame` returns the coding-agent session that produced the code, with citations back to the original transcript and tool calls, alongside fast local search across months of past agent sessions. It launched publicly around July 4, 2026 as open source with a paid 'ctx pro' add-on, and the site states a real limitation: if the session isn't on your machine (e.g., a teammate's agent wrote the code), ctx says it cannot prove attribution. The supplied snippets are all same-author launch material or coverage of it — nothing independently validates the claimed deterministic line-to-tool-call retrieval — and an HN commenter challenged that plain shell search over Claude Code's transcript directories already covers part of this, a point the creator conceded. Peripheral hits (enterprise audit-trail checklists citing SOX/HIPAA, signed tool-call envelope patterns, Microsoft-style post-hoc agent accountability posts) confirm agent-action auditability is a live broader concern but add nothing on ctx's accuracy.
Known: the exact primitive — a reverse walk from a committed line back to the originating agent session — is already carried at book length by Scott's own canon (ip:framework.provenance-coupled-work and ip:source.provenance-coupled-work-ebook specify the conversation×artefact join and reverse walk; ip:source.code-what-transcript-why-ebook already names the ~30-day transcript-deletion timer Ctx's value depends on), so a fifth/sixth indie re-implementation is the world repeating his published argument rather than a consequential party newly arriving — the shell-search concession and paywall friction refine Ctx's weaknesses, not the thesis. Stays medium rather than low because it bears on what he actively builds: the own-machine attribution limit, per-use paywall resistance, and retention-durability dependency are direct design and productization signals for dev:project.search-conversations and dev:project.search, which occupy exactly this territory.
ip:framework.provenance-coupled-workip:source.provenance-coupled-work-ebookip:source.code-what-transcript-why-ebookip:framework.cognitive-gitdev:project.search-conversationsdev:project.searchradar:ctx-agent-session-blameradar:claude-code-30-day-session-deletionradar:claude-code-session-url-metadataradar:pond-s3-agent-session-archiveradar:hindcast-claude-code-session-recovery
queries asked of Scott's wikis
- provenance-coupled work commits agent sessions
- Search Conversations transcript retrieval agent memory
- agent-maintained wiki decision record why code
- coding agent harness transcript logging hooks
- local index RAG over agent conversation history
- design decisions durability transcript retention
now 0 pts/hpeak 0 pts/hcomments 0/hpeers p14momentum: steady2 platformsage 914h
points/hour across evidence · reading as of 2026-10-12 02:59:37.977291+11:00 · deterministic, not a model opinion
2026-09-24T21:55:52Z
grounded: known/medium — Known: the exact primitive — a reverse walk from a committed line back to the originating agent session — is already carried at book length by Scott's own canon
2026-09-24T21:48:54Z
Critic (Feyn) is a second independent product for interrogating agent-authored changes, extending the auditability pattern from static per-commit evidence records to interactive review with the authoring agent — five independent implementations across distinct architectures now support the thesis, while Ctx's own line-to-transcript claim remains same-author and unverified.
2026-09-24T20:37:40Z
evidence attached: hn.story.49834098 — Second independent product for reviewing and interrogating agent-authored changes — pattern-spread evidence for the agent-code auditability case.
2026-09-16T15:38:56Z
The same maintainer now describes exact-tool-call retrieval from commit, file, line, or PR entry points alongside local transcript search, sharpening the evaluation target without independently validating it. This is a more specific account of the existing provenance claim, not a demonstrated capability advance or a reason to escalate attention.
2026-09-16T15:22:38Z
evidence attached: hn.story.49727859 — This is direct supporting coverage of ctx's agent-transcript-to-code provenance workflow and materially strengthens the released-artifact evidence.
2026-09-13T16:25:37Z
Docket adds a headline-level lead for commit-level evidence records, not validation of Ctx’s line-to-transcript linkage. The earlier accelerating label conflated growth in adjacent tooling with corroboration of Ctx itself; this narrower claim remains worth watching but unverified.
2026-09-13T16:22:09Z
evidence attached: hn.story.49685642 — Docket is a related released artifact for binding agent-written code to per-commit evidence, materially reinforcing the code-provenance episode.
2026-09-10T19:00:22Z
The new Claude Code hooks announcement adds an adjacent capture mechanism, not evidence that recorded actions can be reliably traced back from committed code. Independent implementations still support the emerging transcript-provenance pattern, while Ctx’s line-level accuracy, retention durability, and real-world adoption remain unvalidated.
2026-09-10T18:24:04Z
evidence attached: hn.story.49648068 — Tool-call logging provides relevant first-party evidence for auditing agent actions and their resulting code.
2026-09-08T19:42:08Z
No new evidence changes the established interpretation: independent tools support an emerging transcript-provenance pattern, not validation of Ctx’s line-level linkage. Further attention should hinge on demonstrated linkage reliability or real-world use rather than repeated review of the same releases.
2026-09-06T18:28:13Z
The independent implementations still support an emerging transcript-provenance tooling pattern, but this review adds no evidence of adoption or validation of Ctx’s line-level linkage. Preserve that substantive interpretation while lowering monitoring urgency; sibling releases do not establish Ctx’s accuracy or durability.
2026-09-04T17:31:51Z
Gage adds a third independent implementation and extends provenance from retrieving commit-linked intent to automatically inspecting agent transcripts for risks absent from diffs. The emerging tooling pattern is now moving, although Ctx’s specific line-level coverage and the tools’ real-world reliability remain unvalidated.
2026-09-04T17:23:20Z
evidence attached: hn.story.49566640 — Gage independently reinforces the hypothesis that agent transcripts provide auditable evidence missing from code diffs alone.
2026-09-04T14:35:46Z
Casefile supplies an independent implementation of commit-linked agent intent and session history, turning Ctx from an isolated release into evidence of an emerging provenance pattern. It does not independently validate Ctx’s claimed line-level accuracy, durability, or coverage.
2026-09-04T14:22:56Z
evidence attached: hn.story.49564994 — Independent evidence supports the case that linking agent transcripts and task intent to commits can make agent-authored code more auditable and explainable.
2026-09-03T14:40:45Z
No new evidence or engagement corroborates the maintainers’ release claim; the case remains a highly relevant but unvalidated implementation rather than an emerging broader pattern.
2026-09-03T14:31:50Z
grounded: known/high — The radar already tracks this exact development on `radar:ctx-agent-session-blame`. Ctx’s claimed line-to-session linkage closely matches Scott’s published tran
2026-09-03T14:29:50Z
case created — The artifact introduces a specific provenance primitive for tracing production code back to an agent’s reasoning session.