Open-source · SDLC orchestration

Tazuna.

/ ta-zoo-na / · 手綱 · the reins

You don't replace the brilliant agent doing the work. You hold the reins. Tazuna steers AI through stages, approvals, and artifacts — so intelligence reaches its destination.

The metaphor

The reins, not
the rider.

In the Mahabharata, Krishna was Arjuna's sarathi — his charioteer. He didn't fight the war. He steered the most powerful warrior to victory through every stage of the battle.

Modern AI agents are powerful, fast, and capable. What they need is structure: a path through requirements, design, implementation, review, and release — with the right artifacts, tools, and human approvals at each gate.

Tazuna is that path. You define it. The agent runs it. You hold the reins.

The primitive

Track Stage Task

Every workflow in Tazuna is built from three concepts. Compose them once, run them on any project.

01 · Track

The journey

An end-to-end workflow — feature delivery, bug triage, release. Reusable as a template.

02 · Stage

The phase

A bounded phase with a sign-off gate. Requirements, design, implementation, review.

03 · Task

The work

The unit an agent (or human) executes. Inputs, outputs, artifacts, schema.

What's inside

Everything you need to steer.

Tazuna is more than a workflow engine — it's a complete control surface for AI-assisted software development.

AI Agents

Autonomous, collaborative, and manual task runners. Plug in Claude via CLI or API. Bring your own.

Approvals

Human-in-the-loop sign-offs between stages. Review artifacts before the next phase runs.

Artifacts

First-class deliverables — specs, code, designs — linked and versioned across the entire track.

Hooks

Event-driven extensibility. React to task start, stage complete, artifact saved — automate anything.

Integrations

Two-way sync with Jira and Azure DevOps. Tracks map to epics, stages to sprints, tasks to stories.

AI Track Designer

Describe a workflow in plain English. Tazuna proposes a track structure — stages, tasks, schemas.

In 60 seconds

Get started.

Prerequisites:

  • Node.js 20+
  • PostgreSQL 14+
  • Claude Code CLI or an Anthropic API key

Three commands to a running Tazuna. Clone, install, migrate. Open the client at localhost:5173.

Open the repo
~/tazuna
# 1. Clone
$ git clone https://github.com/ga7188/Tazuna.git
$ cd Tazuna

# 2. Server
$ cd server
$ cp .env.example .env
$ npm install
$ npm run migrate
$ npm run dev

# 3. Client (new terminal)
$ cd client
$ npm install && npm run dev

# → http://localhost:5173