OpenCrabs is a new open‑source AI agent that runs entirely on your own hardware. It can read code, fix bugs, and even improve itself over time. In this article we’ll walk through what makes OpenCrabs special, how it works, and why you might want to try it out.

What Is a Self‑Hosted AI Agent?

A self‑hosted AI agent is a program that you install on your own servers or local machine. It does not rely on a cloud service for its core intelligence. Instead, it uses models that you can run locally, and it keeps all data on your own network. This gives you full control over privacy, cost, and customization.

OpenCrabs is built for developers who want a powerful AI assistant without sending data to a third‑party provider. It can run on a laptop, a home server, or a small cloud instance.

The OpenCrabs Architecture

OpenCrabs is written in Rust and uses a text‑based user interface (TUI) that works in the terminal, Slack, Discord, Telegram, and more. Its core is split into three parts:

  1. Brain – The memory that stores all the agent’s knowledge. It keeps a history of conversations, code snippets, and tool usage.
  2. Skills – Small programs that perform specific tasks, such as running a compiler, calling a web API, or analyzing a video file.
  3. Router – The decision engine that chooses which skill to use based on the user’s request.

The brain is persistent, so when you restart OpenCrabs it remembers everything it learned before. The router uses a simple rule‑based system that can be extended with new skills.

Key Features

  • Self‑Improving – OpenCrabs can generate new skills and add them to its skill set automatically. It learns from failures and successes.
  • Self‑Healing – If a skill fails, OpenCrabs can retry with a different approach or fall back to a backup skill.
  • Cross‑Platform – Works on Linux, macOS, and Windows. It can be run inside Docker or as a native binary.
  • Video Vision – The latest release adds a analyze_video tool that can process short video clips using Gemini.
  • OpenRouter Integration – OpenCrabs can call any LLM provider that supports the OpenRouter API, giving you flexibility in choosing models.

How to Install OpenCrabs

Installing OpenCrabs is straightforward. You need a recent Rust toolchain and a terminal. Follow these steps:

  1. Install Rust 
    curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
  2. Clone the Repository 
    git clone https://github.com/adolfousier/opencrabs.git
cd opencrabs
  3. Build the Binary 
    cargo build --release
  4. Run the Agent 
    ./target/release/opencrabs

You can also run it inside Docker:

docker run -it --rm ghcr.io/adolfousier/opencrabs:latest

Once the agent starts, you’ll see a TUI that looks like a chat window. You can type commands or ask questions directly.

Using OpenCrabs in Your Workflow

OpenCrabs can help with many everyday tasks. Below are some common use cases:

1. Code Review and Bug Fixing

Type a command like:

/fix my code

OpenCrabs will read the file, run a static analysis, and suggest fixes. If you have a large codebase, you can point it to a directory and let it scan automatically.

2. Documentation Generation

Ask OpenCrabs to generate documentation for a module:

Article supporting image

/docs src/utils

It will read the source files, extract comments, and produce Markdown docs.

3. Data Analysis

If you have a CSV file, you can ask:

/analyze data.csv

OpenCrabs will load the data, run basic statistics, and give you a summary.

4. Video Analysis

With the new video vision skill, you can analyze a short clip:

/analyze_video path/to/video.mp4

OpenCrabs will send the clip to Gemini and return a description of what’s happening.

Security and Privacy

Because OpenCrabs runs locally, all data stays on your machine. There is no network traffic to external servers unless you explicitly use a skill that calls an API. This makes it ideal for sensitive projects.

OpenCrabs also includes a built‑in security audit skill:

/security-audit

It scans your code for common vulnerabilities and gives you a score out of 100.

Comparison to Other Self‑Hosted Agents

Feature OpenCrabs Other Agents
Language Rust Python, Go
TUI Yes No
Self‑Improving Yes No
Video Vision Yes No
OpenRouter Support Yes Limited
Security Audit Built‑in External tools

OpenCrabs stands out because it combines a lightweight Rust core with powerful AI skills and a self‑learning loop.

Future Roadmap

The OpenCrabs team is actively adding new features:

  • More Built‑in Skills – Adding support for Docker management, Kubernetes, and CI/CD pipelines.
  • Better UI – A web interface is in the works.
  • Community Skill Marketplace – Users can share and download skills.

If you’re interested in contributing, check out the GitHub repo and open an issue.

Why You Should Try OpenCrabs

  • Control – Keep your data on your own hardware.
  • Cost‑Effective – No monthly cloud fees.
  • Extensible – Add new skills or tweak the router.
  • Learning – See how an AI agent can improve itself.

OpenCrabs is a great tool for developers who want a hands‑on AI assistant without the overhead of a cloud service.

Conclusion

OpenCrabs is a powerful self‑hosted AI agent that learns and fixes itself. Its Rust foundation, TUI, and built‑in skills make it a versatile tool for developers. Whether you’re debugging code, generating docs, or analyzing videos, OpenCrabs can help. Because it runs locally, you keep full control over your data and costs. Give it a try and see how it can fit into your workflow.