What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Agent Pigeon is a local, read-only command-line tool that summarizes completed Claude Code and Codex sessions. Its reports surface observed reads and edits, recognized verification runs, failures followed by passes, and session timing—useful clues for reviewing an agent’s work, but not a complete transcript or a grade.
Jay_Stride introduced the tool in a DEV Community post on September 26, 2026. The public CLI was reviewed against its source code and was at v0.1.2 in that post. The author disclosed that AI tools helped draft and edit the post and assisted with parts of the launch workflow. Read the original article.
What Agent Pigeon shows
A coding agent’s final response can omit the activity that produced it: files it touched, checks it ran, or retries it made. Agent Pigeon condenses selected events from completed local session histories so a developer can decide where to inspect the original record.
Depending on the session data and what the parser can recognize, a report may include:
#1 Best Overall
- Observed READ and EDIT activity.
- Recognized verification runs and FAIL→PASS sequences.
- The span between the first and last recorded events, plus the final recognized verification state.
- Repository-documented details such as the largest debugging loop, most-touched file, and longest stretch of coding without recognized verification.
These are observed signals, not a reconstruction of every action. EDIT counts file events, not changed lines or diff size. A FAIL→PASS sequence shows order, but cannot establish that a particular edit caused the later pass. Session span includes idle time between recorded events.
Choose a command for the question you have
| Command | Scope and output | How to interpret it |
|---|---|---|
npx agent-pigeon flight |
A quick report for the latest session; the repository also documents selecting a session by prefix. | Use it to review observed activity in one session. It is not a verdict on the agent. |
replay |
Activity across a corpus of sessions. | Provides a broader activity view, still limited to events the tool can identify. |
compare <A> <B> |
Side-by-side facts for two selected sessions. | Useful for comparing recorded signals; it does not score models or declare a winner. |
share flight or share compare |
SVG output for a flight report or comparison. | Creates a shareable visual summary, not additional evidence about the session. |
For the quick report, run npx agent-pigeon flight in a terminal. The repository documents the other commands and selection options in its README.
Why READ can be N/A instead of zero
Agent Pigeon distinguishes unavailable information from a measured absence. The DEV post explains that Claude Code exposes Read calls directly, while Codex may perform reads inside broader exec commands, making individual reads hard to attribute reliably. In that situation, READ: N/A means the parser cannot count them; READ: 0 would imply it saw the full set and found none.
The choice reflects a trade-off in the tool’s design. The author says incomplete event history caused false positives and noisy warnings in earlier experiments, so the public CLI reads completed sessions in read-only mode rather than presenting live activity.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallWhat the report cannot establish
- Verification detection is heuristic. It may miss custom scripts or checks the parser does not recognize.
- Read attribution may be unavailable. This is why a count can appear as N/A rather than zero.
- Event order is not causation. A failure followed by a pass does not prove which change fixed the issue.
- File events are not change size. EDIT activity does not say how many lines changed or how large a diff was.
- Elapsed span is not active work time. It runs from the first recorded event to the last and includes idle periods.
The Agent Pigeon repository puts the intended use succinctly: “Reports are prompts to inspect, not grades.” Open the underlying session when the summary raises a question; do not treat an unrecognized check or unavailable count as proof that no work occurred.
Install and understand the local-data claims
The repository currently documents Node.js 20.11 or newer and global installation with npm install -g agent-pigeon. It says the CLI reads session histories from ~/.claude/projects and ~/.codex/sessions in read-only mode. According to the repository, it does not read project source code, write state or cache, or transmit data; reports contain aggregate counts and short session IDs rather than source code, diffs, commands, or prompts. These are the project’s stated behaviors, not independently reproduced guarantees. Review the repository and its code if your environment requires an independent privacy or security assessment.
Rank #4
The repository says it has tested Windows 11 and Linux (Debian 12, Node 20, Docker), while macOS is untested. Those platform notes are the project’s own statements. It lists an MIT license.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Who is likely to find it useful
Agent Pigeon is most relevant to developers who already use Claude Code or Codex and want a compact index into completed sessions. It can help identify sessions with recognized checks, repeated failures, or concentrated file activity, then direct a manual review. It is less useful when you need a full audit trail, reliable attribution for every read, a measurement of code quality, or proof that a specific edit caused a test to pass.
Best Value
The article’s demonstration GIF uses sanitized synthetic history, so its displayed counts are illustrative rather than statistics about real agents. Neither the article nor repository presents an independent study or performance benchmark.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
