An AI coding agent built for the terminal - A Go implementation inspired by OpenCode.

• Native TUI: A responsive, native terminal UI built with Bubble Tea
• LSP Enabled: Automatically loads the right language servers for the LLM
• Multi-session: Start multiple agents in parallel on the same project
• Multi-Provider Support: Works with 11+ LLM providers:
  • Anthropic (Claude 3.5 Sonnet/Haiku)
  • OpenAI (GPT-4o, GPT-4o mini)
  • Google Gemini (1.5 Pro/Flash, 2.0)
  • Mistral AI (Large, Small, Codestral)
  • Cohere (Command R+/R)
  • Perplexity (Sonar Large/Small Online)
  • DeepSeek (Chat, Coder)
  • xAI (Grok Beta)
  • Groq (Llama 3.3 70B - Ultra-fast inference ⚡)
  • Together AI (Llama 3.1 405B, Qwen 2.5 Coder, DeepSeek V3)
  • Cerebras (Llama 3.3 70B on Cerebras WSE)
• Local Models Support 🏠:
  • Ollama (localhost:11434) - 100% free & private
  • LM Studio (localhost:1234) - GUI-based local inference
  • Any OpenAI-compatible endpoint
• Agent System: Customizable agents with permission management
• 7 Built-in Tools: read, write, edit, bash, list, grep, glob
• HTTP API Server: RESTful API for integration with other tools

• 🚀 Ultra-fast providers: Groq (300+ tokens/s), Cerebras (hardware acceleration)
• 🏠 Privacy-first: Run 100% locally with Ollama - no data leaves your machine
• ⚡ Instant startup: Go binary starts in milliseconds
• 📦 Single binary: No npm install, no node_modules
• 🎯 Focus on code: Terminal-first design keeps you in flow
• 🔧 Extensible: Add custom providers via simple config

• Go 1.25.4 or higher
• Docker (optional, for containerized development)
• Git

go install github.com/maurorisonho/gocode/cmd/gocode@latest

Or build from source:

git clone https://github.com/maurorisonho/gocode
cd gocode
go build -o gocode ./cmd/gocode

# Set up Groq (ultra-fast, cheap)
export GROQ_API_KEY="gsk_..."

# Start TUI
gocode

# Or run single prompt
gocode run -p groq "Create a REST API in Go"

# Install Ollama
curl -fsSL https://ollama.ai/install.sh | sh

# Pull a model
ollama pull llama3.2

# Start Ollama server
ollama serve

# Configure gocode
cat > ~/.gocode/config.yaml << EOF
llm:
  default_provider: ollama
  default_model: llama3.2

providers:
  ollama:
    base_url: http://localhost:11434/v1
    enabled: true
EOF

# Use it!
gocode

# Fast agent with Groq
gocode run -p groq -m llama-3.3-70b-versatile "quick task"

# Complex task with Claude
gocode run -p anthropic -m claude-3-5-sonnet-20241022 "refactor code"

# Private task with local model
gocode run -p ollama -m llama3.2 "analyze sensitive code"

See docs/PROVIDERS.md for complete setup guide.

Execute a prompt directly without the TUI:

gocode run "Explain how to use context in Go"

# With specific provider and model
gocode run -p groq -m llama-3.3-70b-versatile "Fast response"
gocode run -p ollama -m llama3.2 "Private task"

Start as a background server:

gocode serve --port 4096

GoCode looks for configuration in .gocode/ directory in your project root or in ~/.config/gocode/.

Create .gocode/config.yaml:

llm:
  default_provider: groq # Ultra-fast!
  default_model: llama-3.3-70b-versatile

providers:
  # Fast inference
  groq:
    api_key: ${GROQ_API_KEY}
    enabled: true

  # Best quality
  anthropic:
    api_key: ${ANTHROPIC_API_KEY}
    default_model: claude-3-5-sonnet-20241022
  openai:
    api_key: ${OPENAI_API_KEY}
    default_model: gpt-4o
  gemini:
    api_key: ${GEMINI_API_KEY}
    default_model: gemini-1.5-pro

agents:
  build:
    description: "Full development agent with all tools"
    tools:
      bash: allow
      write: allow
      edit: allow

  plan:
    description: "Planning agent without file modifications"
    tools:
      bash: ask
      write: deny
      edit: deny

permissions:
  edit: ask
  bash:
    "*": ask
    "git *": allow
    "ls *": allow