ORG-2: A Session Recorder and Replay System for AI Coding Agents
System of record for how agents build software: Built-in rust harness & 20+ CLIs.
At a glance
- What is it?
- ORG-2 is a Tauri desktop app that records every AI coding agent session as a replayable trajectory, capturing messages, tool calls, file edits, and command output in a synchronized timeline. It fills the gap between git blame, which traces committed lines, and the agent decisions that produced those lines.
- Who is it for?
- ORG-2 is worth evaluating for teams running multiple AI coding agents who need a shared record of how code was produced. The trajectory replay and AI blame features address a real gap in current tooling: git shows the what, ORG-2 shows the why and the how.
- Can I use it commercially?
- Yes, with strict conditions. AGPL-3.0 is a network copyleft licence: if people use a modified version over a network, for example as a hosted service, you must offer them its source code under the same licence.
- Is it still maintained?
- Yes. The repository received new commits within the last day.
- What is it written in?
- Mainly TypeScript, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on September 30, 2026, and from our analysis. They are not legal advice.
Editorial analysis
The Recording Gap That ORG-2 Fills
When humans write code, tools like Jira, GitHub, and Amplitude cover different slices of the process. Jira sees tickets; GitHub sees commits; Amplitude sees runtime metrics. None of them see the decision process between ticket creation and the final diff. At human speed, that gap is survivable. The README makes the trade-off explicit: 'Code written on Monday is legacy by Friday' at agent speed.
ORG-2 is designed for teams that have moved to AI coding agents and need to understand not just what code was produced, but which session produced it, what the agent understood from the human's request, and what sequence of tool calls led to a given line. The README describes the project as 'an experiment in human/agent organizations and org-level alignment' that treats agents as persistent, observable colleagues inside a structured organization rather than as tools that produce opaque output.
Session Trajectories: Replaying Agent Work Like a Video
The core feature of ORG-2 is trajectory replay. When a coding agent session runs through the built-in Rust harness or through a supported CLI, ORG-2 records messages, tool calls, file edits, and command output in a synchronized timeline. Teammates can replay this timeline as if watching a video of the session, reviewing the actual sequence of decisions rather than only the resulting diff.
Sessions from external tools are ingested and backfilled from their history, so the record covers work that never ran inside ORG-2 directly. The README notes that the record links what the human asked for, what the agent understood, and what it actually did. This makes it possible to review an agent's reasoning path before approving a pull request, rather than only checking the output.
Team review is built into the workflow: sessions can be shared across devices and teammates, and comments can be left in context at specific points in the trajectory, not just at the diff level.
Downloading and Installing ORG-2
ORG-2 ships as a native desktop application built with Rust and Tauri. The README states it runs under 100MB on disk. Downloads are provided for four platforms directly from the latest GitHub release:
- macOS Apple Silicon: `ORG2-latest-mac-apple-silicon.dmg` - Windows x64: `ORG2-latest-windows-x64-setup.exe` (installer) or `ORG2-latest-windows-x64.msi` (MSI for enterprise deployment) - Linux x64: `ORG2-latest-linux-x64.AppImage` (portable) or `ORG2-latest-linux-x64.deb` (Debian package)
The README does not document a macOS Intel build in the listed release assets. After installing, ORG-2 runs locally; the README describes it as local-first. Cross-device session sharing and team collaboration use a self-hosted Supabase backend, which the README lists as a work-in-progress feature.
The project uses TypeScript for the frontend, Rust for the Tauri harness, and pnpm as the package manager. The package.json version at the time of the last push was 2.0.8; the latest release at that point was v2.0.9, published on 2026-09-27.
AI Blame: Tracing Committed Lines to Agent Sessions
Git blame answers who changed a line and when. ORG-2 adds what the README calls AI blame: the ability to trace a committed line back to the specific agent session, tool calls, and decisions that produced it.
This matters in practice because multiple agent sessions may touch the same file over time. Git blame shows the last author; AI blame shows the session that generated the code, the prompt that drove it, and the intermediate tool calls between the human request and the final write. For debugging regressions or understanding why a piece of code behaves unexpectedly, knowing the originating session gives more context than knowing the committer.
The README positions AI blame alongside full dev workspace tooling: terminal access, Git history, and pull request review are all available without leaving the agent workspace. The Design Mode feature adds a native WebKit browser for inspecting live pages, with the ability to select an element and send its context directly to the active agent.
The Built-in Rust Harness and the 20+ Supported CLIs
ORG-2 includes its own coding agent implemented as a native Rust harness. The README describes it as fast, token-saving, and customizable, using the user's own API keys and agent subscriptions rather than requiring a separate account. This harness records trajectories in the same format as sessions from external tools, so all replay and review features work identically.
For teams that use other agents, ORG-2 supports launching and managing sessions from more than twenty agent CLIs, including Claude Code, Cursor CLI, OpenAI Codex, GitHub Copilot, Kiro CLI, OpenCode, DeepSeek Harness, and others listed in the README under Supported Agents. Sessions from these tools can be scanned, inspected for subagents, and controlled from inside ORG-2 without switching applications.
The README also documents cross-session memory, cross-agent knowledge sharing, resource-aware execution that can react to CPU and RAM availability, and scheduling for overnight or unattended sessions. These features are described as capabilities in the current release, though the README notes that org-level alignment surfaces for issues and projects management are still a work in progress.
AGPL-3.0 License Implications and Real Limitations
ORG-2 is released under AGPL-3.0. This license permits free use, modification, and redistribution, but it requires that modified versions made available over a network also publish their source code. For commercial teams that want to modify ORG-2 and deploy it internally as a service, this is a meaningful constraint that differs from permissive licenses like MIT or Apache-2.0. Teams with legal requirements around license obligations should review the AGPL-3.0 terms before integrating ORG-2 into a commercial workflow.
A practical limitation is that the session collaboration and group issue workflow features depend on self-hosted Supabase, which the README marks as work in progress. Teams that need multi-user session sharing today should test whether this feature is fully operational in the current release.
A reasonable alternative for teams that do not need trajectory replay is to use standard PR review tooling with detailed commit messages that reference the agent and the prompt. GitHub's code review interface is widely available, handles diffs, and supports inline comments. What it does not do is show tool call sequences or agent reasoning steps, which is where ORG-2's replay feature provides distinct value.
The last push was on 2026-09-25 and the latest release was published on 2026-09-27, which places ORG-2 under active development at the time of writing.
Editorial conclusion
ORG-2 is worth evaluating for teams running multiple AI coding agents who need a shared record of how code was produced. The trajectory replay and AI blame features address a real gap in current tooling: git shows the what, ORG-2 shows the why and the how. The AGPL-3.0 license is a significant constraint for commercial teams that deploy modified versions in a networked context; review the license before integrating ORG-2 into a commercial workflow. The self-hosted Supabase collaboration features are listed as work in progress, so teams depending on multi-user session sharing should verify their status before committing to the platform.
Frequently asked questions
Does ORG-2 work with Claude Code and other coding agents I already use?
The README lists over twenty supported CLIs, including Claude Code, Cursor CLI, OpenAI Codex, GitHub Copilot, Kiro CLI, and OpenCode. Sessions from these tools can be loaded, scanned, and managed inside ORG-2. The built-in Rust harness is also available as an alternative to external CLIs.
What does AI blame show that git blame does not?
Git blame identifies the last committer and commit for a line. AI blame, as described in the README, traces a committed line back to the agent session, the specific tool calls, and the human decisions that drove the change. This is intended for situations where understanding why the code was written matters as much as knowing who committed it.
What are the commercial use restrictions for ORG-2?
ORG-2 is licensed under AGPL-3.0. Teams that modify ORG-2 and make it available over a network must publish the modified source code. Unmodified use and private internal use do not require source publication, but deploying a modified version as a networked service does. The license file in the repository contains the full terms.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/org2ai-org2)