Garda Studio
Contact
Command-line interface · v1.4.3

Available · Open source

A local-first governance layer for AI coding agents.

Garda Agent Orchestrator 1.4.3 adds guarded custom review lanes and authenticated remediation to a strict, repeatable cycle from task entry to verified completion.

Install Garda CLI View on GitHub

2-minute demo

Garda in action.

See how a task moves through review findings, focused fixes, and verified completion.

Recorded walkthrough · English captions available · Watch on YouTube

Local dashboard / garda ui

Every task, gate and review in one view.

Run garda ui to open a read-only dashboard on 127.0.0.1. Follow a task through its stages, check reviews and full-suite results, and see the profiles and quality rules your workflow uses.

Garda UI task queue for the Garda repository: task counters, navigation tabs and the task table with status, priority, area and title
Garda's own task queue: Garda is developed through Garda.
Task progress in Garda UI: completed stages with timestamps, the current stage and the remaining stages
Task progress: what passed, what is running, what is still ahead.
Latest quality check in Garda UI: evidence current, status PASS, changed files and rule answers marked PASS
Latest quality check: evidence, changed files and every rule answer.

Workflow / Before and after

What changes when every agent follows one workflow.

Without Garda

  • The agent jumps straight into edits.
  • Tests, reviews, and docs checks depend on discipline.
  • Completion means whatever the agent claims is done.
  • There is no shared audit trail for another human to inspect.

With Garda

  • Supported provider surfaces follow one canonical task workflow.
  • Preflight classifies scope and risk before implementation.
  • Fresh delegated reviews account for every assigned coverage obligation.
  • The task ends with findings receipts, gate evidence, and tamper-evident history.

Task profiles

Tailored task profiles.

Profile

Fast

For small low-risk tasks that should stay lightweight.

Profile

Balanced

The default mode for day-to-day development.

Profile

Strict

For risky changes, deeper review, and stronger guarantees.

Profile

Docs-only

For documentation work without unnecessary code-review overhead.

Runtime / Core capabilities

Core capabilities.

Govern many agents

Use popular providers and local provider surfaces without letting each agent invent its own workflow.

Configure review lanes

Add guarded custom lanes, bind profile states and dependencies, and freeze the effective catalog for each task cycle.

Remediate with proof

Garda chooses authenticated REUSE, bounded DELTA, or conservative FULL review from current evidence and scope.

Follow next-step

garda next-step routes catalog dependencies, remediation, recovery, and closeout through the exact next command.

Start with the process: How to build a coding-agent workflow with checks and independent review.

Download

Get Garda Agent Orchestrator

Garda Agent Orchestrator is published on npm. Install it globally and run it inside any repository.

Status
Available
Version
1.4.3
Platforms
Node.js 22.13+ or 24+
npm install -g garda-agent-orchestrator

Support Garda Agent Orchestrator

Buy me some tokens

Garda Agent Orchestrator is an independent open-source project for governed AI coding workflows. Voluntary contributions support this project specifically, not the other Garda Studio products.

Support on Ko-fiDonate crypto

What support helps cover

  • AI tokens
  • Model subscriptions
  • Release validation
  • Documentation
  • Compatibility testing
  • Maintenance

Support is voluntary. It is not a Garda AI Desktop subscription and does not guarantee private support, roadmap priority, or custom development.