Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsRexo Code is an open-source AI coding agent written in Rust and operated from the terminal. Its developer, Daksh Saboo, describes it in a DEV Community article published September 27, 2026 as a provider-agnostic tool that works on a project through a series of tool-using steps, not a single prompt sent to a model. This article explains what the project does, how its agent loop is designed, what the author says about building it, and which details you should confirm before installing a current release.
What Rexo Code is
Rexo Code is a command-line tool that puts a language model inside a working loop with your code. According to the author’s article and the project’s README, it is built to understand and search a project, read and edit files, run shell commands, and then check what happened before deciding its next move. Read the list below as the capabilities the project reports, not as features that have been independently tested.
As an Amazon Associate I earn from qualifying purchases.
- Understanding and searching project files
- Reading and editing files
- Running shell commands and inspecting their results
- Using MCP tools
- Asking for permission before sensitive actions
- Managing persistent sessions
- Using skills and custom commands
- Running hooks and plugins
- Working with subagents
- Accepting image input
- Streaming responses
- Running in headless mode with JSON output for automation
The project is free and open source, according to the article. Its repository describes it as actively developed.
How the agent loop works
The clearest difference between Rexo Code and a simple chat wrapper is the loop. The repository describes each request moving through the following stages:
#1 Best Overall
- Request: the user states a task in the terminal.
- Project context: the agent gathers information about the project it is working in.
- Model reasoning: the selected model decides how to approach the task.
- Tool selection: the model chooses a tool, such as a file read, an edit, or a shell command.
- Permission check: sensitive operations are checked against the permission rules before they run.
- Tool execution: the approved tool runs.
- Verification or next action: the agent inspects the result and either verifies the outcome or moves to the next step.
According to the README, this cycle repeats until the task is complete, the agent needs input from the user, or it reaches an execution boundary it is not allowed to cross. The author’s broader point is the same: an agent has to act on real files and commands, so the surrounding machinery matters as much as the model.
Why the author says it is not just an API wrapper
The article’s central argument is that the job is different from building an application that calls a model API. In the author’s words: “Building an AI coding agent is quite different from just making an application that sends a prompt to an API.” (Daksh Saboo, “I Built an AI Coding Agent in Rust,” DEV Community)
That difference shows up in the list of concerns the project has to handle: project context, tool use, permission boundaries, command output, session state, and extension points. Each adds a place where the agent can go wrong, which is why the design centers on the loop rather than on the single model call.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Rank #2
Choosing a provider
The author says the project was designed so users are not tied to one AI provider, and can choose models and services that fit their workflow. The README lists these options:
- NVIDIA NIM
- OpenAI-compatible APIs
- Google Gemini
- Local model servers
- Custom OpenAI-compatible endpoints
Being listed in the README does not mean each provider has been individually tested for this project, and the list is not an endorsement of any of them. If you use a local model server or a custom endpoint, check that it implements the interface the project expects before relying on it for real edits.
Permissions and verification
Because the agent edits files and runs commands, the author treats permissions as a core design concern. The repository says operations can require approval before execution, and the loop includes an explicit step where the agent inspects command results and decides whether to continue. Together, these give you two checkpoints: a gate before an action runs, and a review after it runs.
Rank #3
Neither checkpoint guarantees a safe outcome on its own. A permission prompt tells you what the agent wants to do; it does not tell you whether the change is correct. Review diffs and test results yourself, especially for changes that touch several files.
How the author built it
The article says the author used Claude and Claude Code for debugging, implementation, refactoring, and larger changes across the Rust codebase. The author also says they tested the changes and decided what belonged in the project. In other words, the AI tools were assistants in the work, and the author remained responsible for verification and scope.
The development process is described as iterative: build, test, hit breakage, diagnose the problem, and rebuild. That cycle is ordinary for a project of this kind, and the article’s value is in showing where the hard parts were, namely the agent loop, permissions, and provider handling, rather than in any claim of speed or productivity.
Versions and platforms
Two version labels appear in the sources, and they do not match. The article identifies v0.9 as the release the author felt comfortable offering more widely. When the repository was reviewed, its interface displayed 9.0.0. The sources do not establish how these labels map to each other, so do not treat either one as the definitive current version. Check the repository’s releases page for the current build.
| Item | Source | What the source says |
|---|---|---|
| Supported operating systems | Article | Windows, Linux, and macOS |
| Prebuilt targets | README | Windows x86_64, Windows ARM64, Linux x86_64, macOS ARM64 |
| macOS signing status | README | Releases are currently unsigned and not notarized, which may require additional approval before the program runs |
| Version label | Article | v0.9, described as the release the author felt comfortable offering more widely |
| Version label | Repository interface, when reviewed | 9.0.0 |
The macOS status is the detail most likely to change. If you use a Mac, check the current release notes before installing, and follow the repository’s instructions for approving an unsigned application.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Installing and running it
The repository offers two routes: downloading a prebuilt release, or building from source with Cargo. It describes interactive use in the terminal and headless use that produces JSON output for scripts and automation. This article does not provide a full installation walkthrough, and the commands and prerequisites change between releases, so follow the current instructions in the Rexo-Code repository on GitHub.
What is and is not established
The article and the README are the only sources behind the claims above. Neither includes an independent benchmark, a measurement of developer productivity, or a third-party security review of Rexo Code. Performance, safety, and reliability statements should therefore be read as the author’s and the project’s own descriptions. Where this article describes a feature, it reflects what the project says it does; if you need to rely on a specific behavior, test it in your own environment first.
For readers interested in the broader design question, the most useful part of the project is the explicit loop: request, context, reasoning, tool choice, permission, execution, and verification. Any agent you build or evaluate should be judged on how it handles each of those stages.
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.
Recommended Free Tools
