taracode

DevOps & Cloud AI Assistant
Expert guidance for Kubernetes, Terraform, Docker, and multi-cloud deployments.
Runs locally with Ollama for complete privacy.

Why taracode?

• DevOps Expertise - Specialized in Kubernetes, Terraform, Docker, CI/CD, and cloud platforms
• 58 Built-in Tools - DevOps, security scanning, file operations, git, web search
• Multi-Agent System - 7 specialized agents for complex tasks
• Privacy-first - Runs fully local with Ollama, your data never leaves your machine
• No Account Required - Open source, just install and use

Quick Start

1. Install Ollama

# macOS / Linux
brew install ollama

# Or download from https://ollama.ai

2. Pull a Model

ollama pull gemma4:12b     # 16 GB machines
ollama pull qwen3.8:27b    # 32 GB machines (needs Ollama 0.32.12 or newer)
ollama pull qwen3.6:35b    # 48 GB and up

Any model that Ollama tags with the tools capability works. Ollama's default context window is 4,096 tokens
on machines with less than 24 GB of GPU memory, which is too small for the tool schemas; start Ollama with
OLLAMA_CONTEXT_LENGTH=32768 (or raise Context length in the Ollama app settings). taracode warns you when the
server's window is too small.

3. Install taracode

Quick install (recommended):

curl -fsSL https://code.tara.vision/install.sh | bash

Homebrew (macOS / Linux):

brew install --cask tara-vision/taracode/taracode

Upgrading from a version installed as a formula? Run brew uninstall taracode once, then the command above.

Go install:

go install github.com/tara-vision/taracode@latest

Manual download:

Download binaries from GitHub Releases. Every release ships checksums, a cosign signature and SLSA provenance; see SECURITY.md to verify a download.

4. Run

cd your-project
taracode
> /init    # Initialize project features

That's it! Start asking questions about your infrastructure.

Features

Screen Monitoring (/watch)

Let the AI watch your screen and catch errors before you do:

> /watch this          # Capture and analyze all screens now
> /watch start         # Start continuous monitoring
> /watch stop          # Stop monitoring

Multi-Agent System

7 specialized agents work together on complex tasks:

> /agent list          # List all agents
> /agent use security  # Route next prompt to specific agent

Autonomous Task Execution (/task)

Plan and execute multi-step tasks with checkpoints:

> /task "Add authentication to the API"
> /task "Deploy to production with blue-green strategy"
> /task templates      # List built-in templates

Project Memory

Remember project-specific knowledge across sessions:

> /remember We use PostgreSQL for production databases
> /remember Always run tests before pushing #workflow
> /memory search database

DevOps Tools

Security Mode

Full DevSecOps capabilities with audit logging:

> /mode security       # Switch to security mode

# Security scanning
> Scan this image for vulnerabilities: nginx:latest
> Check for secrets in the current directory
> Run a SAST scan on the codebase

Commands

Configuration

Create ~/.taracode/config.yaml:

# Single Host (simple setup)
host: http://localhost:11434

# Multi-Host Setup (v2.0) - for multiple Ollama servers
hosts:
  primary:
    url: http://gpu-server:11434
    models: [ qwen3.8:27b, gemma4:12b ]
    priority: 1
  local:
    url: http://localhost:11434
    fallback: primary      # Use primary if local is down
    priority: 2
default_host: primary

# Model generation options (v2.0.4)
model:
  temperature: 0.7     # Sampling randomness (0.0-2.0)
  top_p: 0.9           # Nucleus sampling threshold (0.0-1.0)
  num_predict: 0       # Max tokens per response (0 = model default)

# Search
search:
  primary: duckduckgo
  fallback: searxng
  brave_api_key: ""    # Optional: Brave Search API

# Memory
memory:
  enabled: true
  auto_capture: true

# Per-agent host assignment
agents:
  coder:
    host: primary
    model: qwen3.8:27b
  reviewer:
    host: local
    model: gemma4:e4b

See config.example.yaml for all options.

Supported LLM Backends

Roadmap

v3 turns taracode into the local-first DevOps operator: read-only investigation by default, policy-gated
operations, a smaller tool set that fits small context windows, a native Ollama client, a published eval
scoreboard of local models, and an MCP server plus skills pack. See ROADMAP.md.

Development

make deps      # Install dependencies
make build     # Build binary
make test      # Run tests
make install   # Install to /usr/local/bin

See CONTRIBUTING.md for development guidelines.

Contributing

Contributions are welcome! Please read our Contributing Guide
and Code of Conduct.

Security

For security issues, please see our Security Policy.

License

MIT License - see LICENSE for details.

Built with ❤️ by Tara Vision · Created by Dejan Stefanoski