claude-code-Go

README

claude-code-Go — AI Coding Agent in Go

A Go-native AI coding agent with full agent loop, tool execution, permission management, SSE streaming, and auto-recovery — in a single binary.

Status: v0.3 verified release — Core agent runtime, permission system, session persistence, multi-provider support, doctor checks, replay evidence, MCP/LSP extension diagnostics, and manifest-driven harness quality gates are implemented and tested. Real external MCP/LSP server smoke checks and competitor-agent comparisons remain manual. See PARITY.md for detailed feature status.

Disclaimer: This is an independent open-source project. It is not affiliated with, endorsed by, or connected to Anthropic PBC. "Claude" and "Claude Code" are trademarks of Anthropic PBC.

Installation

Via go install (Recommended)

go install github.com/strings77wzq/claude-code-Go/cmd/go-code@latest

The binary installs to $GOPATH/bin (typically ~/go/bin). Make sure it's in your PATH:

export PATH="$HOME/go/bin:$PATH"

Build from Source

git clone https://github.com/strings77wzq/claude-code-Go.git
cd claude-code-Go
make install
go-code

Pre-built Binaries

Download from GitHub Releases:

# Example: Linux amd64
curl -fsSL https://github.com/strings77wzq/claude-code-Go/releases/latest/download/go-code-linux-amd64 -o go-code
chmod +x go-code
sudo mv go-code /usr/local/bin/

For script-based install, see scripts/ for install.sh and install.ps1.

Quick Start

1. Set your API key

# Anthropic
export ANTHROPIC_API_KEY=sk-ant-...

# Or Tencent Cloud Coding Plan
export ANTHROPIC_API_KEY=sk-sp-...
export ANTHROPIC_BASE_URL=https://api.lkeap.cloud.tencent.com/coding/anthropic
export ANTHROPIC_MODEL=tc-code-latest

Or create ~/.go-code/settings.json:

{
  "apiKey": "sk-ant-...",
  "model": "claude-sonnet-4-6-20251001"
}

2. Verify your setup

Run the health check before starting a real session:

go-code doctor

For offline environments:

go-code doctor --offline

3. Run

# Interactive TUI
go-code

# Legacy REPL
go-code --legacy-repl

# Single prompt (non-interactive)
go-code -p "Explain the agent loop architecture"

# JSON output (for scripting)
go-code -p "List files in current directory" -f json

# Replay a saved session
go-code replay latest

# Collect concise release/issue evidence
go-code replay --evidence latest

Verified Features (v0.3)

These features are tested and covered by the parity harness. See PARITY.md for evidence links.

Feature	Status	Tests
Agent Loop (think → act → observe)	Verified	Go unit + harness
11 Built-in Tools	Verified	Go tests
Permission System (3-tier)	Verified	Go + harness
Doctor Health Check	Verified	Go tests
Multi-Provider (Anthropic, OpenAI-compatible)	Verified	Go tests
Session Persistence + Resume	Verified	Go tests
Session Replay + Evidence Mode	Partial v0.3	Go tests + harness
Slash Commands (/help, /model, /sessions, etc.)	Verified	Go tests
Skills System	Verified	Go tests
Hooks System	Verified	Go tests
MCP Integration	Partial v0.3	Go tests + harness
LSP Integration	Partial v0.3	Go tests + harness

Built-in Tools

Read, Write, Edit, Glob, Grep, Bash, Diff, Tree, WebFetch, TodoWrite, NotebookEdit

Supported Providers

Provider	Setup
Anthropic	ANTHROPIC_API_KEY=..., optional LLM_PROVIDER=anthropic
OpenAI	LLM_PROVIDER=openai, ANTHROPIC_API_KEY=..., ANTHROPIC_BASE_URL=https://api.openai.com
DeepSeek	LLM_PROVIDER=openai, ANTHROPIC_BASE_URL=https://api.deepseek.com, model deepseek-chat or deepseek-reasoner
Qwen	LLM_PROVIDER=openai, ANTHROPIC_BASE_URL=https://dashscope.aliyuncs.com/compatible-mode
GLM	LLM_PROVIDER=openai, ANTHROPIC_BASE_URL=https://open.bigmodel.cn/api/paas
Tencent Cloud	LLM_PROVIDER=anthropic, ANTHROPIC_BASE_URL=https://api.lkeap.cloud.tencent.com/coding/anthropic, model tc-code-latest

Architecture

┌─────────────────────────────────────────────────────────────┐
│                    claude-code-Go                            │
├─────────────────────────────────────────────────────────────┤
│  ┌──────────────┐    ┌──────────────┐    ┌───────────────┐  │
│  │  Bubbletea   │───▶│ Agent Loop   │───▶│ Tool Registry │  │
│  │     TUI      │    │   + Context  │    │               │  │
│  └──────────────┘    └──────────────┘    └───────────────┘  │
│                            │                    │            │
│                            ▼                    ▼            │
│                     ┌──────────────┐    ┌───────────────┐  │
│                     │   API Client │    │   Built-in    │  │
│                     │ (SSE Stream) │    │    Tools      │  │
│                     └──────────────┘    └───────────────┘  │
└─────────────────────────────────────────────────────────────┘

For a detailed architecture overview, see docs/architecture/.

Project Structure

claude-code-Go/
├── cmd/go-code/          # Main entry point
├── internal/
│   ├── agent/            # Agent loop + context management
│   ├── config/           # Multi-source config loader
│   ├── cost/             # Cost tracking and estimation
│   ├── hooks/            # Pre/post execution hooks
│   ├── lsp/              # Language Server Protocol client
│   ├── permission/       # 3-tier permission system
│   ├── provider/         # Multi-provider abstraction
│   │   ├── anthropic/    # Anthropic API provider
│   │   ├── openai/       # OpenAI-compatible provider
│   │   └── registry/     # Provider auto-selection
│   ├── session/          # Session persistence + resume
│   ├── skills/           # Custom skills system
│   ├── tool/             # Tool interface + builtins
│   │   ├── builtin/      # 11 built-in tools
│   │   ├── mcp/          # MCP integration
│   │   └── init/         # Tool registration
│   └── update/           # Auto-update checker
├── pkg/
│   ├── tty/              # Legacy REPL (use --legacy-repl)
│   └── tui/              # Bubbletea TUI (default)
├── harness/              # Python test harness
├── docs/                 # VitePress documentation
├── scripts/              # Install and launch scripts
└── .github/workflows/    # CI/CD

Development

make build          # Build binary
make test           # Go + Python harness tests
make vet            # Static analysis
make build-all      # Cross-compile all platforms
make docs           # Serve docs locally
make docs-build     # Build docs for production

Documentation

Full documentation: https://strings77wzq.github.io/claude-code-Go/

Key pages:

• Quick Start
• Architecture Overview
• Roadmap
• Troubleshooting
• MCP Integration
• LSP Integration
• Contributing
• Parity Status

License

MIT License — see LICENSE for details.