Centaurus

The Unified Engineering Terminal.

One terminal. Your entire workflow.

Code. Deploy. Debug. Operate. Automate.

All from one place.

Why Centaurus Exists

Engineering work is fragmented.

Developers constantly switch between:

• terminals
• SSH sessions
• containers
• deployment tooling
• runbooks
• AI assistants

Centaurus unifies them.

One Input. Everything.

$ git status
On branch main, your branch is up to date with 'origin/main'.

> Find where auth middleware is used.
→ src/controllers/auth.ts:142 · src/services/session.ts:68 · src/middleware/auth.ts:34

$ docker logs api
[INFO] API server listening on port 8080

$ ssh production
Connected to [email protected]

> Investigate the memory spike.
→ Checking memory usage...
→ Found: Redis connection pool not releasing idle connections
→ Suggested fix: Add pool maxIdle and idleTimeoutMillis configuration

/workflow deploy-validation
✓ typecheck: tsc --noEmit
✓ test: vitest run --reporter=dot
✓ build: next build
✓ push image: docker push registry/api:sha
✓ notify team: posted to #deployments

Core Differentiators

Environment Awareness

Centaurus knows where it is. SSH into a remote server, enter a Docker container, switch to WSL — the AI detects the transition automatically, runs tools in that environment, and keeps the conversation intact.

Intent-Aware Terminal

One input. Everything. Commands, AI requests, workflow invocations, and environment operations all share the same interface. The system automatically determines intent without mode switching.

Persistent Workspace

One workspace. Many tasks. Run development servers, monitor logs, maintain SSH sessions, and delegate AI work without opening additional terminals. Everything stays running while you continue working.

Workflow Engine

Workflows are engineering processes, not shell scripts. A workflow can span local machines, SSH sessions, Docker containers, WSL environments, and multiple servers. Define once, run with a single command.

Autonomous Execution

Turn Centaurus loose. Proactive mode keeps the AI running between turns — monitoring, building, investigating — without waiting for you to prompt it. Parallel agents work on separate tasks while you work on yours.

Model Freedom

Centaurus routes to Google Vertex AI, Anthropic, AWS Bedrock, OpenRouter, NVIDIA NIM, Silicon Flow, local Ollama, LM Studio, and custom endpoints. Switch models with /models. The conversation continues.

Precision Control

Four auto-accept modes let you tune exactly how much autonomy Centaurus has. model-decides is unique: the AI itself signals whether an action is safe to auto-approve. Every turn is checkpointed. /revert undoes everything.

Install

curl -fsSL https://centauruslabs.in/install.sh | sh

Quick Start

# Launch Centaurus
centaurus

# Initialize your project
/init

# SSH into production and investigate
ssh production
> Investigate why the API is slow

# Run a cross-environment workflow
/workflow deploy-staging

Keep Everything Running

Centaurus is not merely a command runner. It is a persistent engineering workspace.

Frontend Dev Server      Running (localhost:3000)
Backend API             Running (localhost:8080)
Production Logs         Streaming (ssh prod)
Integration Tests       Running (background task 2)
AI Investigation        Running (sub-agent)

Dev servers, logs, tests, and AI agents stay active while you continue working.

Documentation

• Full documentation
• Slash command reference
• Workflow guide
• Skills and automation
• Environment awareness