Trending GitHub Repos: 4 Open Source AI Coding Agents

9 min read

Came here from Instagram, LinkedIn, X or YouTube? Here is the full source list you came for. Everything below is built from these, nothing else.

The leaderboard

• GitHub Trending, weekly view, checked Oct 24 2026

The repos

• morluto/rea, README and rea.tools, the project site
• debpalash/VoiceStudio, README
• stablyai/orca, README
• anomalyco/opencode, README, OpenCode docs and CONTRIBUTING.md

The release notes

• rea-agents v6.7.0, Oct 18 2026
• VoiceStudio v0.5.6, Sep 23 2026
• Orca v1.4.239, Oct 24 2026
• OpenCode v1.19.4, Oct 22 2026

The receipts

• Creation dates, stars, forks and licenses from the public GitHub API (api.github.com/repos/owner/name), checked Oct 24 2026

On October 24, 2026, I opened GitHub's weekly trending page and found one repo that had picked up 88,243 stars in seven days. That repo is rea. It now sits at 375,326 stars and 93,838 forks.

The next three repos by stars gained that week were VoiceStudio with 12,526, Orca with 9,545 and OpenCode with 6,765. All four are tools for AI agents.

Here is the part that will annoy you. Three of those four repos did not exist at the start of this year. Orca was created on March 17, 2026. VoiceStudio on April 9. rea on April 14. Only OpenCode is older, created on April 30, 2025.

So the usual reading of a trending page, "these are the big established projects everyone uses", is wrong here. Most of this list is six or seven months old. And the trending page does not sort by stars gained either, so OpenCode shows up at position 5 even though it gained more stars than every repo above it except rea.

This post is a tour. What each project does, who it is for, the exact command to try it, and where you could send your first pull request.

Key takeaways

• rea gained 88,243 stars in the week, about 23.5 percent of its 375,326 total. It is an MCP server and CLI that lets your coding agent reverse engineer apps and binaries.
• VoiceStudio gained 12,526 stars. It is a local, open source alternative to ElevenLabs under AGPL-3.0.
• Orca gained 9,545 stars. It runs a fleet of coding agents side by side, each in its own git worktree.
• OpenCode gained 6,765 stars and is the largest of the group after rea, at 220,888 stars. It is an MIT licensed coding agent with a build mode and a read only plan mode.
• Three of the four repos were created between March 17 and April 14, 2026.
• Star counts move every minute. Re-check them on the day you read this.

What is trending on GitHub this week?

Here is the snapshot I took. The weekly star counts and totals come from the trending page. Creation dates and licenses come from the GitHub API.

Repo	What it is	Stars this week	Total stars	Created	License	Source
morluto/rea	Agent-driven reverse engineering	88,243	375,326	Apr 14, 2026	MIT	Trending page, GitHub API
debpalash/VoiceStudio	Local ElevenLabs alternative	12,526	84,006	Apr 9, 2026	AGPL-3.0	Trending page, GitHub API
stablyai/orca	Orchestrator for coding agents	9,545	104,337	Mar 17, 2026	MIT	Trending page, GitHub API
anomalyco/opencode	Open source coding agent	6,765	220,888	Apr 30, 2025	MIT	Trending page, GitHub API

Two things stand out when you put them side by side.

First, rea is not just ahead. It took about 75 percent of the stars these four repos gained together that week (88,243 of 117,079).

Second, growth is not the same as size. OpenCode gained the fewest stars of the four, but 6,765 on top of 220,888 is about 3 percent growth in a week for a project that has been around for 18 months. rea grew by almost a quarter of its total in the same week.

The rest of the weekly list follows the same theme. Anthropic's Claude Code plugin directory, Cursor's plugin spec, a memory layer for agent CLIs, and HeyGen's "Write HTML. Render video. Built for agents." all sit in the top 15. Even the NSA's Ghidra is there with 4,221 new stars, which fits, because rea can drive it.

What does rea do, and who is it for?

The rea README describes it in one line: "Reverse engineer anything with agents, from app behavior down to native binaries."

In practice it is an MCP server plus a CLI. Your agent (the README names Claude Code, Codex, Cursor, Gemini CLI and Grok Build) calls rea to inspect a target. rea returns findings with evidence and limitations, and the agent uses them to explain the behavior or write and test an implementation. Analysis runs locally.

The list of targets is long: native binaries through Hopper, Ghidra or IDA, Go, Rust and .NET binaries, JavaScript and Electron apps, websites, Android APKs, firmware, packet captures, SQLite databases, Unity IL2CPP metadata, EVM bytecode and PyInstaller executables. Static JavaScript and .NET inspection need no native engine at all.

The project site shows the kind of question it is built for. One example prompt is "Use REA to inspect Windows Calculator. Why does 200 + 10% give 220?" Another asks why Chrome's dinosaur game gets faster, then asks for a small version with adjustable speed.

Try it:

npx rea-agents setup

That configures your agent. To run an analysis straight from the terminal on an extracted Electron or JavaScript app:

npx -y rea-agents@latest analyze-javascript-application /absolute/path/to/app --json

You need Node.js 22.x (22.23.4 or later), 24.x (24.22.1 or later) or 26.12.1 and up, plus whatever target tools your analysis needs, such as Ghidra, adb or tshark.

Two cautions. rea's own code is MIT, but the README says commercial tools like IDA Pro, Hopper and JEB must be legally obtained by you. And only point it at software you have the right to inspect. The latest release, v6.7.0 on October 18, also notes that the folded-in 6.6.0 changes break saved Evidence and snapshots, so regenerate those if you were already using it.

Is VoiceStudio a real ElevenLabs alternative?

For running voice work on your own machine, that is exactly how it describes itself: an open source, local first alternative to ElevenLabs for "voice cloning, voice design, video dubbing, dictation, transcription & audiobook creation in 646 languages." Remote services are optional.

The default engine is powered by k2-fsa/OmniVoice, and you can switch to other engines. It also exposes a local API and MCP support, which is why it belongs in an agent tools roundup. The v0.5.6 release on September 23 added setup guides for connecting Claude Code, Cursor, Codex CLI and the OpenAI Agents SDK.

Hardware matters here, so check this before you install:

• NVIDIA GPU on Windows or Linux: CUDA acceleration.
• Apple Silicon: Metal acceleration.
• No dedicated GPU: runs on CPU, more slowly, and needs about 5 GB of free disk for the CPU build of PyTorch.
• Intel Mac: UI only, connecting to a remote backend.

Try it on macOS or Linux:

curl -fsSL https://voicestudio.sh/install | sh

On Windows, in PowerShell:

irm https://voicestudio.sh/install | iex

Read the install script before you pipe it into a shell. That is true for every curl-to-sh line in this post.

Three things to know before you build on it. The app is AGPL-3.0, which has real consequences if you ship a modified version as a network service. Model weights, voices and generated output have their own separate license terms. And the README tells you to clone voices only with permission. The release notes also say the Electron installers are unsigned or ad hoc signed, so expect trust warnings on Windows and macOS.

How does Orca manage a fleet of coding agents?

Orca's README calls it "The AI Orchestrator for 100x builders." Strip the slogan and the idea is simple: you run several coding agents at once, each in its own git worktree, and you watch them all from one window.

It works with any CLI agent. The README lists Claude Code, Codex, Cursor, GitHub Copilot, OpenCode, Amp, Goose, Cline, Kiro and Qwen Code among others.

The features that matter day to day:

1. Parallel worktrees. Send one prompt to several agents, compare what each produced, merge the best one.
2. Annotate AI diffs. Comment on specific diff lines and send the comments back to the agent.
3. SSH worktrees. Run agents on a remote machine with auto reconnect and port forwarding.
4. Design mode. Click an element in an embedded Chromium window and its HTML, CSS and a screenshot go to the agent.
5. Mobile companion. iOS and Android apps notify you when an agent finishes and let you send follow-ups.

Try it on macOS:

brew install --cask stablyai/orca/orca

On Arch Linux:

yay -S stably-orca-bin

Windows and other Linux builds are on the GitHub releases page.

Orca ships fast. The latest release, v1.4.239, landed on October 24, the same day I checked the trending page. It added one click installs of Orca's agent skills, per-repo browser profiles and better SSH recovery. The repo also had 8,677 open issues and pull requests on the GitHub API that day, so if you contribute, expect a busy queue.

Why is OpenCode still trending a year and a half in?

OpenCode is the steady one. The README calls it "The open source AI coding agent", and it is MIT licensed. The docs say it runs as a terminal interface, a desktop app or an IDE extension, and that you can use "any LLM provider by configuring their API keys."

It has two built-in agents you switch between with the Tab key:

• build, the default, with full access for development work.
• plan, read only. It denies file edits by default and asks permission before running bash commands.

That plan mode is the reason I would suggest OpenCode to someone nervous about letting an agent loose on a real codebase. Start in plan, read what it proposes, then switch.

Try it:

curl -fsSL https://opencode.ai/v2/install | bash

Or through npm:

npm i -g @opencode/cli@latest

Then, inside your project, run /connect to pick a provider and /init. The docs say /init analyzes your project and creates an AGENTS.md file in the root, which you should commit so the agent understands your structure and patterns.

The latest release, v1.19.4 on October 22, is a small one: it bumps the bundled Bun runtime to 1.4.4 for security fixes and updates root certificates.

Which one should you try or contribute to?

Pick by the problem you have this week, not by the star count.

If you want to	Try	First command
Understand how an app or binary works	rea	npx rea-agents setup
Clone, dub or transcribe voice locally	VoiceStudio	curl -fsSL https://voicestudio.sh/install | sh
Run several agents in parallel on one repo	Orca	brew install --cask stablyai/orca/orca
Use one open source agent with any model	OpenCode	curl -fsSL https://opencode.ai/v2/install | bash

If you want to contribute, OpenCode has the clearest path of the four, because its CONTRIBUTING.md spells it out. It welcomes bug fixes, extra LSPs and formatters, LLM performance improvements, environment specific fixes and documentation. New features and UI changes need a design-approved issue first, and bug fix PRs must reference an existing issue with Fixes #123. Look for labels like good first issue and help wanted, and comment on the issue to claim it.

The local setup needs Bun 1.3 or newer:

bun install
bun dev

Run tests and type checks from the package you changed, not the repo root:

cd packages/core
bun run test
bun typecheck

For VoiceStudio, the README gives a from-source path that doubles as a contributor setup:

git clone https://github.com/debpalash/VoiceStudio.git
cd VoiceStudio
bun install
bun run setup:api
bun run dev

My honest advice: do not open a PR on any of these in your first hour. Install the tool, use it on something real, and file the bug you hit. A good bug report on a repo moving this fast is worth more than a drive-by typo fix.

Actionable takeaways

• Install one of the four today and run it on a real project, not a demo.
• Before you run any curl | sh installer, open the script and read it.
• If you use VoiceStudio in a product, read AGPL-3.0 and the separate model and voice terms first.
• If you use rea, only inspect software you have the right to inspect, and get any commercial disassembler legally.
• For OpenCode, start in plan mode and commit the AGENTS.md that /init creates.
• Re-check the star counts on the day you share this. They will have changed.

FAQ

What are the trending GitHub repos for AI coding agents right now?

On GitHub's weekly trending page on October 24, 2026, the four repos with the most stars gained were rea (88,243), VoiceStudio (12,526), Orca (9,545) and OpenCode (6,765). Other agent tools in the top 15 included Anthropic's Claude Code plugin directory, Cursor's plugins and ai-memory. The page updates constantly, so check it yourself for today's list.

Is rea legal to use?

rea's own code is MIT licensed. Its README says commercial tools such as IDA Pro, Hopper and JEB must be legally obtained by the user. Whether a specific analysis is allowed depends on the software you inspect and its license, so only use it on targets you have the right to examine.

Does VoiceStudio need a GPU?

No. It runs on the CPU, more slowly, and needs about 5 GB of free disk for the CPU build of PyTorch. NVIDIA GPUs get CUDA acceleration on Windows and Linux, Apple Silicon gets Metal acceleration, and Intel Macs only run the UI against a remote backend.

What is the difference between Orca and OpenCode?

OpenCode is a coding agent: it reads your code and makes changes. Orca is an orchestrator: it runs many agents, including OpenCode, side by side in separate git worktrees so you can compare and merge their work. You can use them together.

Pick one of the four and install it before the end of the day, then browse the free tools on Art of Code for the small jobs around it. If you are looking for work on teams building with these tools, check the Art of Code jobs board.

Written by Adithya Guttha, Founder of Art of Code.