---
type: "post"
id: "thruline-0-3-0-ships-mac-app-that-bdd58c"
slug: "thruline-0-3-0-ships-mac-app-that-bdd58c"
title: "ThruLine 0.3.0 ships Mac app that checks coding-agent 'done' claims against local run logs"
description: "ThruLine’s 0.3.0 Mac app checks coding-agent claims against locally recorded session logs, including what ran and on which code version."
retrieval_nugget: "ThruLine’s 0.3.0 Mac app checks coding-agent claims against locally recorded session logs, including what ran and on which code version. Its value is a deterministic evidence trail before approval—not a verdict on whether the code is correct—and its narrow Mac-only scope makes it an early tool to evaluate rather than a general workflow standard."
published_at: "2026-11-07"
updated_at: "2026-11-07"
record_date: "2026-11-07"
date_kind: "published_at"
topics: ["claude","claude-code","codex"]
entities: ["ThruLine","Claude Code","Codex","RAAS Labs","DSSE","in-toto"]
related_patterns: ["verification-bandwidth-is-the-scarce-resource"]
novelty_label: "new_fact"
what_changed: "ThruLine 0.3.0 adds a Mac app that checks coding agents’ completion claims against local run logs and whether the relevant code was re-tested."
prior_refs: ["https://newruntime.com/posts/opencode-team-ships-hex-3-0-a-stable-ef277c/","https://newruntime.com/posts/release-notes-delta-docs-6253bf/"]
source_url: "https://thrulinehq.com/"
source_urls: ["https://thrulinehq.com/"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:thruline-0-3-0-ships-mac-app-that-bdd58c"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/thruline-0-3-0-ships-mac-app-that-bdd58c/","markdown":"https://newruntime.com/posts/thruline-0-3-0-ships-mac-app-that-bdd58c.md","json":"https://newruntime.com/posts/thruline-0-3-0-ships-mac-app-that-bdd58c.json"}
markdown_url: "https://newruntime.com/posts/thruline-0-3-0-ships-mac-app-that-bdd58c.md"
---

# ThruLine 0.3.0 ships Mac app that checks coding-agent 'done' claims against local run logs

## Retrieval answer

ThruLine’s 0.3.0 Mac app checks coding-agent claims against locally recorded session logs, including what ran and on which code version. Its value is a deterministic evidence trail before approval—not a verdict on whether the code is correct—and its narrow Mac-only scope makes it an early tool to evaluate rather than a general workflow standard.

## Summary
ThruLine’s 0.3.0 Mac app checks coding-agent claims against locally recorded session logs, including what ran and on which code version. Its value is a deterministic evidence trail before approval—not a verdict on whether the code is correct—and its narrow Mac-only scope makes it an early tool to evaluate rather than a general workflow standard.

## What happened
ThruLine’s product page describes version 0.3.0 as a Mac app for checking an AI coding agent’s “done” claims against what actually ran. It is available for Apple silicon Macs running macOS 12 or later. The app is signed and notarized by Apple, installs from a disk image, and updates itself after asking. The page identifies Claude Code and Codex as supported contexts for its examples, and says the product is not affiliated with Anthropic or OpenAI.

The app reads the session record an agent already writes locally. For each claim, it shows the claim, the run cited as evidence, the code version on which that run occurred, and whether relevant files changed afterward. The product illustrates this with an agent claiming that tests pass: the cited run reports four passed and zero failed, but it ran on a version older than the one under consideration, and a file changed after the run. ThruLine therefore labels the claim “Not checked” on the current version rather than treating the older run as proof. The product says this does not mean the claim is false; it means the record does not establish it for that version.

A workflow called a “Change” breaks acceptance into six stages: define what counts as done, prepare the exact project and starting version, inspect changes, verify, review, and approve. Proposed requirements must be accepted, edited, or dropped before they count. Verification runs a specified command once for the project; outcomes include passed, failed, ran but checked nothing, cancelled, interrupted, or tool not found. Approval applies to one named version. ThruLine says it does not merge, deploy, or tell the agent anything.

The company says the verdict path is deterministic and does not use a model to decide. Records are append-only and hash-chained; the product can produce a signed receipt using a DSSE envelope and an in-toto statement. Its sample receipt comes from a synthetic Python test project and a sample key, not a production identity. ThruLine says records, runs, and receipts stay on the user’s disk and that the core engine makes no network calls. The page notes a daily GitHub version check, which users can disable, plus network activity initiated by users for licence requests or Docker image downloads when running `thruline reexec`.

Five checks a month are free. Pro costs $15 per month or $150 per year for one developer and adds unlimited checks and approval of a version. The page says recording, reading, and checking the agent’s own runs are free. It says the product is built by one engineer; multi-developer licences require contacting the company. Version 0.3.0 is closed source, while version 0.2.0 is listed as public under the FSL-1.1-MIT licence.

## What the archive adds
New Runtime’s earlier coverage of DeepSeek’s dsh-superagent described a coding agent taking tasks from approved team-chat requests and working against a GitHub repository ([6 November 2026 note](https://newruntime.com/posts/deepseek-open-sources-super-agent-a-coding-agent-2c70ef/)). That is related by coding-agent workflow, but not the same event: ThruLine adds a local verification product, not another coding agent.

## Trend signal
This supports the view that verification bandwidth is a scarce engineering resource: ThruLine makes run evidence and version scope explicit before approval. The connection is early and product-level; the page does not show that teams have adopted it or reduced review queues ([verification bandwidth is the scarce engineering resource](https://newruntime.com/patterns/verification-bandwidth-is-the-scarce-resource/)).

## Why it matters
Coding agents can produce plausible completion messages that are difficult to verify from a diff alone. ThruLine focuses on a narrower question: does the recorded work support the agent’s claim for the exact version a developer is about to accept? Showing the run, its commit, and later changes can make a missing check visible before approval, while a signed receipt offers a portable record of what was tested.

That boundary matters because “not checked” is not the same as “failed,” and a successful run on an older version is not evidence about newer code. Teams could use this distinction to make acceptance criteria more explicit and to reduce accidental reliance on conversational confidence. But the product currently targets Apple silicon Macs and one developer per Pro licence, so it does not yet establish a team-wide verification workflow or solve review capacity on its own.

## Limits and uncertainty
ThruLine’s examples and claims are published by the company; the sample receipt is from a synthetic project and uses a sample key, not a production identity. The page does not establish how well the product handles varied repositories, test setups, or real-world agent sessions, nor whether its record captures every relevant action. It also does not show independent measurements of reduced defects, faster review, or improved release outcomes. Version 0.3.0 is closed source, limiting external inspection of its implementation.

## Reader takeaway
ThruLine turns an agent’s completion claim into a question about evidence on a specific version: what ran, what it covered, and what changed afterward. That is a useful verification boundary to consider even if the Mac-only product is not a fit. A passing test on an earlier commit is not proof for the code you are approving now.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Claude](https://newruntime.com/topics/claude.md): Explore the claude topic hub.
- topic hub: [Claude Code](https://newruntime.com/topics/claude-code.md): Explore the claude-code topic hub.
- topic hub: [Codex](https://newruntime.com/topics/codex.md): Explore the codex topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)