deepseek-reasonix
DeepSeek-native AI coding agent for your terminal. Engineered around prefix-cache stability — leave it running.
What is it?
What it is
Reasonix is an open-source, config-driven AI coding agent built in Go that can run autonomously in a terminal (TUI), desktop app, browser, or editor via ACP. It supports multi-model providers (including DeepSeek presets and any OpenAI-compatible endpoint), plugin-driven MCP servers, cache-aware context maintenance, and per-turn checkpoints for long-running sessions.
Why it exists
Reasonix exists to provide a single, self-contained local engine for autonomous coding tasks that developers can leave running. It addresses the need for a flexible, extensible, and trustworthy agent with plan mode, permissions, workspace sandboxing, and undo capabilities across multiple interfaces.
Who should use it
Developers seeking an autonomous coding agent with multi-model support and plugin-driven architectureTeams needing CLI/TUI/desktop/browser integration for local AI workflowsUsers requiring config-driven customization of AI agents and toolsOrganizations prioritizing self-contained binaries with zero dependencies
Who should avoid it
Users needing web-based cloud solutions without local deploymentBeginners unfamiliar with Go toolchains and binary compilationProjects requiring pre-built model marketplacesNon-technical users avoiding configuration file management
How it works
A quick walkthrough in plain English
How DeepSeek-Reasonix works in 4 steps
Step 1 of 4
Something triggers the flow
A schedule, webhook, or manual click tells DeepSeek-Reasonix to start.
Features
Advantages
- Open source under MIT license
- Single binary, no additional runtime dependencies
- Cross-platform support (Windows, macOS, Linux)
- Easy installation via npm, Homebrew, or prebuilt archives
- Extensible through plugins and extensions
- Supports multiple LLM providers and model composition
- Cache-aware context management improves long-running autonomy
- Active community and bilingual Discord support
Disadvantages
- Building from source requires Go 1.25+ and additional toolchain for desktop
- Desktop app build needs Node 24+, pnpm 10, and Wails CLI
- Limited to platforms with prebuilt binaries (amd64, arm64)
- Configuration required for providers and models
- Potential complexity for new users
- No built-in cloud hosting; self-hosted only
Installation
cloud
Not mentioned in README
docker
Not mentioned in README
native
npm i -g reasonix # any OS; pulls the prebuilt native binary brew install esengine/reasonix/reasonix # macOS
compose
Not mentioned in README
FAQ
How do I install Reasonix for the CLI/TUI, desktop app, or VS Code extension?
For CLI/TUI, run `npm i -g reasonix` (any OS) or `brew install esengine/reasonix/reasonix` on macOS. For the desktop app, download the installer from https://reasonix.io/?download=desktop#start (macOS .dmg/.zip, Windows .exe/.zip, Linux .deb/.tar.gz). For VS Code, first install the CLI (Path A), then install the extension from the Visual Studio Marketplace (`SivanLiu.reasonix-agent`) or Open VSX Registry. The extension will start your local `reasonix acp` backend automatically.
What is the basic workflow to configure and run Reasonix in an interactive session?
Run `reasonix setup` to configure a provider and model (you’ll be prompted for API keys and model selection). Then start an interactive session with `reasonix`. Inside the session, use `/init` to let Reasonix create project instructions, and you can issue commands like `reasonix run "implement the TODOs in main.go"`. The agent will operate in plan mode, respect permissions, and keep checkpoints for safe, reversible execution.
How does Reasonix handle context, checkpoints, and safe long‑running autonomous runs?
Reasonix injects a small stable environment summary at startup, prunes stale tool output before compacting context, and creates per‑turn checkpoints. This lets you read, rewind, or undo actions during a long autonomous run. Checkpoints are managed via the built‑in checkpoint system (see the CHECKPOINTS guide) and can be inspected or rolled back using the UI or CLI commands.
Can I extend Reasonix with custom tools or plugins, and how do I do it?
Yes. Reasonix is plugin‑driven: MCP servers contribute tools, prompts, and resources. You can declare MCP servers in `reasonix.toml` under `[mcp]`. Additionally, Extension Protocol v1 sidecars can intercept runtime events, contribute providers, and ship versioned plugin packages. Refer to the EXTENSIONS.md and PLUGIN_PACKAGES.md docs for detailed manifest and development instructions.
How do I build Reasonix from source for my platform?
Clone the repo, cd into it, and run `make build` to produce `bin/reasonix` (or `bin/reasonix.exe` on Windows). For cross‑platform binaries, run `make cross` which creates a `dist/` folder with darwin/linux/windows × amd64/arm64 static binaries. The CLI requires Go 1.25+; the desktop build additionally needs Node 24+, pnpm 10, and the Wails CLI (see `make wails-install`). Follow the desktop README for platform‑specific webview dependencies.
What is ACP and how does Reasonix integrate with editors via ACP?
ACP (Agent‑Controlled Protocol) is Reasonix’s editor integration layer. The VS Code extension uses ACP to start a local `reasonix acp` backend, providing native chat, editor context, tool‑call approvals, model selection, and workspace sessions directly in the editor. You can find the ACP guide in `docs/ACP.md` and the extension source at https://github.com/SivanCola/reasonix-vscode.
Featured in Videos
YouTube tutorials and walkthroughs for deepseek-reasonix
Alternatives
Similar projects ranked by category, topics, and text overlap.
