Debugging, replaying, and comparing complex LLM agent trajectories without manual reconstruction. It is catalogued under LLM eval & observability on PulseGate.

Who should use retrial?

retrial is an open-source project built for AI developers and researchers.

Does retrial have a free plan?

Yes — retrial is open source under the MIT license and free to use.

What platforms does retrial run on?

retrial runs on the command line.

Is retrial still maintained?

The GitHub repository shows 20 commits in the last 90 days.

What projects are similar to retrial?

Similar projects tracked by PulseGate include ase-python, constraintloop, and OpenCode Data.

retrial is developed by ArcKansupada.

Back to the index

retrial

Name: retrial
Author: ArcKansupada

PyPIInfrastructure

Retrial is a tool that brings Git-style operations (branch, diff, bisect) to LLM agent trajectories. It records and versions the execution paths of autonomous agents, allowing developers to inspect, replay, and compare runs with actual re-execution. Primarily useful for debugging complex agent behaviors and improving observability in AI agent development.

Open SourceMITCLI

Visit PyPI

Overview

5 features

retrial is a LLM eval & observability project. Debugging, replaying, and comparing complex LLM agent trajectories without manual reconstruction. retrial is an open-source project aimed at AI developers and researchers. The project is open source (MIT). It ships for the command line.

ArcKansupada builds and maintains retrial, and it first shipped in 2026. Development happens publicly on GitHub with 20 commits in the last 90 days. Among its 5 catalogued features are Trajectory Versioning, Agent Replay, and Diff Support.

Summary written by a language model from the project’s public pages.

Trajectory Versioning
Agent Replay
Diff Support
Bisect Debugging
Re-execution

Built with & integrations

AI providers

anthropic

Runs on

CLI

Trust & compliance

License

MIT

Verified signals

HTTPSOpen Source GitHubActive maintenance

Indexing history

What PulseGate has recorded for this listing

Indexed1 Aug · 03:14 UTC
retrial verified against its public source
Source: PulseGate · Open

Frequently asked questions about retrial

What is retrial?: Debugging, replaying, and comparing complex LLM agent trajectories without manual reconstruction. It is catalogued under LLM eval & observability on PulseGate.
Who should use retrial?: retrial is an open-source project built for AI developers and researchers.
Does retrial have a free plan?: Yes — retrial is open source under the MIT license and free to use.
What platforms does retrial run on?: retrial runs on the command line.
Is retrial still maintained?: The GitHub repository shows 20 commits in the last 90 days.
What projects are similar to retrial?: Similar projects tracked by PulseGate include ase-python, constraintloop, and OpenCode Data.ase-python constraintloop OpenCode Data
Who makes retrial?: retrial is developed by ArcKansupada.
When did retrial launch?: retrial first shipped in 2026.

At a glance

Platforms: Cli
Languages: English
Open source: Yes (GitHub)
License: MIT
First seen: 1 Aug 2026
Built for: AI developers and researchers
Model: Open source
Solves: Debugging, replaying, and comparing complex LLM agent trajectories without manual reconstruction.

Registered as

GitHub: ArcKansupada/retrial
PyPI: retrial

Developer

ArcKansupada

Solo developer

GitHub

Open source

View on GitHub

Stars: 0
Forks: 0
Open issues: 0
Last commit: 1 Aug 2026
Commits 90d: 20
Contributors: 1
Authorship: Solo
Default branch: main

Live coverage

Identity confidence: Low · 64
Indexed: 1 Aug 2026
Lifecycle: Alive
First seen: Aug 2026
Last seen: 1 Aug 2026

Identity audit (13)

Slug: retrial-pypi-org
Lifecycle state recorded: 1 Aug 2026
Verification state: Indexed for public listing
Listing state: Listed: yes
Index status: Included in index
Latest evidence snapshot: 1 Aug 2026
Timeline basis: Indexed-at chronology (no inferred launch/funding milestones).
Name from: Written by a language model from the project's public pages.
Category from: Assigned by a language model.
Summary from: Written by a language model from public pages.
Languages from: Detected by a language model from page content.
Last updated: 4 Aug 2026
Canonical URL: https://pypi.org/project/retrial

Ship this? Send a correction — no account, and you get a link to follow it.

Also in LLM eval & observability

Same category — not a similarity match