Who is DeepSeek-Reasonix for?

Developers comfortable with the terminal who want a configurable coding agent on DeepSeek or another OpenAI-compatible model, and who know they need an API key for a provider.

What can you build with DeepSeek-Reasonix?

Ask an agent to implement the TODOs in a codebase from your terminal. Chat with a coding agent inside VS Code, with tool-call approvals, through the ACP backend. Point the agent at a DeepSeek or other OpenAI-compatible model by editing a config file. Extend the agent with MCP servers or plugin packages that add your own tools. Rewind a long autonomous session to an earlier checkpoint when a step goes wrong.

esengine /

DeepSeek-Reasonix

A single-binary coding agent with plan mode, sandboxing and undoable checkpoints

35.9k stars
203 stars since 3 Oct
Go
Updated today
Some experience
Coding agents & dev tools

Open on GitHub (opens in a new tab) Website (opens in a new tab)

Reasonix is a coding agent for larger software tasks. You give it a goal, such as implementing the TODOs in a file, and it works through the code using tools. One local engine sits behind every front end: a terminal UI, the Studio desktop app, a browser, and editors that speak ACP, such as the VS Code extension.

Long autonomous runs can be hard to follow, so Reasonix adds guardrails. Plan mode, permissions, a workspace sandbox and per-turn checkpoints are meant to keep a run readable and let you rewind it. Providers, the agent, enabled tools and plugins are all declared in a reasonix.toml file.

DeepSeek comes as a preset, and any OpenAI-compatible endpoint can be added as a config entry. You can also run two models together, one as planner and one as executor. Tools come from MCP servers or Extension Protocol sidecars. It ships as one static binary. There are two lines: 1.x is stable, and 2.x (Studio) is a fast-moving pre-release.

What it does

Config-driven providers
Models, the agent, tools and plugins are declared in reasonix.toml. DeepSeek is a preset, and any OpenAI-compatible endpoint is a config entry rather than new code.
Planner and executor models
You can run two models together in separate sessions, kept cache-stable so prompt caching keeps working.
Safety and undo
Plan mode, permissions, a workspace sandbox and per-turn checkpoints let you review a long run and rewind it.
Plugins and MCP
MCP servers add tools, prompts and resources. Extension Protocol v1 sidecars can intercept runtime events, add providers and contribute structured UI.
Single static binary
Built with CGO_ENABLED=0, it cross-compiles to six targets with one command and needs nothing else installed on the target machine.

What you could build

Ask an agent to implement the TODOs in a codebase from your terminal.
Chat with a coding agent inside VS Code, with tool-call approvals, through the ACP backend.
Point the agent at a DeepSeek or other OpenAI-compatible model by editing a config file.
Extend the agent with MCP servers or plugin packages that add your own tools.
Rewind a long autonomous session to an earlier checkpoint when a step goes wrong.

Get started

Install the stable 1.x CLI with npm or Homebrew, or download Studio 2.x from the releases page. Run setup to choose a provider and model, then start a session or give it a one-off task.

npm i -g reasonix

reasonix setup

reasonix run "implement the TODOs in main.go"

Commands are copied from the project's README. Read it on GitHub for the full setup.