OrcaReplay

OrcaReplay is a command-line tool that records and replays AI coding agent runs for developers who need to find out why an agent failed.

What it is good for

  • Finding out why a coding agent deleted or broke a file
  • Re-running a failed agent session without paying for model calls
  • Comparing how different AI models handle the same coding task
  • Debugging an agent run after the terminal is closed

At a glance

Who it is for
Setup difficulty
Technical ยท Guided command-line setup

Level 4 of 5. You copy and paste commands into a terminal by following the instructions. Comfortable territory for vibe coders.

Cost
Free and open source
Type
Agent app
Tasks
Works with
Claude Code, Codex CLI, Anthropic, OpenAI, OrcaRouter

How to get started with OrcaReplay

  1. Install the package globally with npm i -g orcareplay, which adds the orca command.
  2. Run orca quickstart to try a sample project and recording with no model, key or tokens needed.
  3. Record your own agent with orca record claude.
  4. Replay the last run offline with orca replay last.
  5. Fork a run onto another model from a given step with orca replay last --from 4.

From the project's own documentation. If a step does not match what you see, the official pages have the latest instructions.

Before you try it

  • Recording a real agent needs the agent installed, a key, a network connection and real tokens.
  • Replay runs recorded tool calls for real, and it is not a sandbox.
  • Tools that open their own network connections can still reach the network during replay.

Momentum

Rank this week
#60
Best rank
#60 week 46, 2026
Stars gained last week
+364
GitHub stars
808

GitHub numbers as of Nov 16, 2026. Updated weekly.

209 531 852 Sep 24, 2026 Nov 16, 2026
GitHub stars for OrcaReplay: 253 on Sep 24, 2026, 808 on Nov 16, 2026. Source: GitHub.

Ranking history

First ranked in week 46, 2026.

Questions

Is OrcaReplay free?
Yes. It is free and open source (Apache-2.0 license).
Do I need to know how to code to use OrcaReplay?
You do not need to write code, but setup means copying and pasting commands into a terminal by following its instructions.
What does OrcaReplay work with?
According to its own documentation, it works with Claude Code, Codex CLI, Anthropic, OpenAI, and OrcaRouter.
How do I get started with OrcaReplay?
1. Install the package globally with npm i -g orcareplay, which adds the orca command. 2. Run orca quickstart to try a sample project and recording with no model, key or tokens needed. 3. Record your own agent with orca record claude. 4. Replay the last run offline with orca replay last. 5. Fork a run onto another model from a given step with orca replay last --from 4.
What should I know before trying OrcaReplay?
Recording a real agent needs the agent installed, a key, a network connection and real tokens. Replay runs recorded tool calls for real, and it is not a sandbox. Tools that open their own network connections can still reach the network during replay.
What is OrcaReplay good for?
Finding out why a coding agent deleted or broke a file, re-running a failed agent session without paying for model calls, comparing how different AI models handle the same coding task, and debugging an agent run after the terminal is closed.

Similar agents to OrcaReplay

  • destructive_command_guard

    Blocks dangerous commands before they run instead of recording and replaying runs afterward.

  • rtk

    Shrinks command output to cut token costs rather than recording agent runs for debugging.

  • Claude Squad

    Runs several coding agents in parallel workspaces instead of replaying a single recorded run.

  • Crush

    Is a terminal coding agent that writes code itself, not a debugging tool for other agents.

  • opencode

    opencode is an open source coding agent that writes and changes code in your project for developers.

  • Claw Code

    Claw Code is a command-line coding agent that runs prompts and code tasks from a terminal for developers.

Also compared with OrcaReplay on these pages:

Last updated Nov 16, 2026. Tracked since Oct 29, 2026. Spotted an error? Tell us.