The desktop app is herev0.0.2

OPEN SOURCE · DESKTOP · TERMINAL

Your next idea.
Your next commit.

A focused workspace for coding with AI.
Choose your model. Explore your project. Build, then verify.

Available for macOS, Windows, and Linux

Deepzora/deepzora-workspace
Add HTTP retry handling
Add retry handling to the HTTP client. Explore the current implementation first.
Deepzora

Understand first. Then build.

I’ll read the request logic and existing tests, identify which failures are safe to retry, and propose an implementation plan.

Read project context
src/client.tssrc/client.test.tspackage.json
Explore relevant files
What would you like to work on next?
Your model Ask
Local workspacemain
CODE & CHANGES
client.ts
TypeScriptUTF-8
Interactive illustration · Explore the workflow

BUILT FOR THE WORK ITSELF

Less switching.
More moving forward.

From understanding a project to reviewing a change, keep your tools in step with your thinking.

Think with your whole project

Let the agent read relevant files, edit code, and run commands and tests. Keep conversations, project files, and diffs in one workspace.

Project context

Bring the model you prefer

Use DeepSeek presets or connect an OpenAI-compatible or Anthropic-compatible provider. Configure the connection with your own API key.

Your model connection

Stay part of every step

Review a plan before implementation. Choose tool permissions, inspect the diff, and rewind changes made by supported file-editing tools.

Plans, permissions & checkpoints

KEEP THE WAY YOU LIKE TO WORK

One agent. More ways in.

Explore in the desktop app, move quickly in the terminal, or connect an ACP-compatible editor.

TERMINAL & AUTOMATION

Your terminal. A new teammate.

Code interactively or run a focused task with a single command. The CLI and desktop app share your model configuration.

Read the CLI guide
Terminal

# Install, then start your first task

$ npm install -g @sivancola/deepzora
$ deepzora setup
$ deepzora

$ deepzora run "Fix the failing tests and explain the changes"

MAKE IT YOUR OWN

Keep your tools. Extend your flow.

Connect tools through MCP, add reusable skills and extensions, or bring the agent into an ACP-compatible editor.

Explore integrations
⌘MCPTools & services
✳SkillsReusable workflows
ACPEditor integration

START HERE

Make room for your next idea.

Choose your platform and architecture to download an official release.

Current version v0.0.2

Choose Apple Silicon for M-series Macs, or Intel for Intel-based Macs.

After installation, connect your model provider and API key to get started. See the release notes for signing status and installation instructions.

A FEW THINGS TO KNOW

Before you get started.

What is Deepzora?

Deepzora is an independent open-source AI coding tool derived from Reasonix 1.x, with its own desktop app, CLI, brand, maintenance, and release channels. It is not an official Reasonix edition and is not part of the Reasonix Studio 2.x release line.

Do I need my own model API key?

Connect your own model provider and configure an API key where required. DeepSeek presets and custom OpenAI-compatible and Anthropic-compatible connections are supported. Model usage is billed by your chosen provider.

Where do my code and conversations go?

Application settings, credentials, and conversation history are stored on the device running the agent. Task context is sent to your configured model provider; enabled tools and integrations may also access relevant data. See the privacy policy for details.

Can I review and undo changes?

You can inspect plans, tool calls, and diffs, and choose tool permissions. Checkpoints can rewind changes from supported file-editing tools; shell-command side effects are outside rewind coverage.

Can I use the source or contribute?

Deepzora is distributed under AGPL-3.0-only. You can use, modify, and distribute it subject to the applicable licenses. Original permissions for inherited components remain in effect. Find the source, licensing notes, and contribution guide on GitHub.

DOCS & RESOURCES

A little guidance goes a long way.

From connecting a model to finishing your first task, the docs help you find your footing.