CCode AI

An original agentic coding assistant CLI for local projects with an improved terminal TUI, slash commands, sessions, animated thinking states, and safe coding-agent tools.

Repository: https://github.com/anmolpromax8-creator/Claude-code

Not affiliated with Anthropic and not the official Claude Code product.

Install

npm i -g ccode-ai

Then run:

ccode
# or
ccode-ai

What's new

• Improved terminal UI with splash/header, status frame, compact non-wrapping lines, and input bar

• Ctrl+O expands/collapses the input bar

• Animated generating indicator while the model responds

• Assistant output appears character-by-character in the terminal

• Stable non-animated thinking block disappears before the final response, instead of flickering or leaving “generating” text behind

• Markdown tables are rendered as compact terminal tables

• NVIDIA NIM provider support using the official base URL only:

  • https://integrate.api.nvidia.com/v1

• /model auto-fetches the available NVIDIA NIM model list from the official /models endpoint and lets you pick one

• If the selected provider API key is missing, the CLI asks for it securely

• API keys are saved locally in ~/.ccode/keys.json with 0600 permissions after you enter them once

• /logout forgets the saved API key for the current provider

• More stable compact TUI rendering for narrow/mobile terminals

• Safer mobile terminal rendering avoids full-width auto-wrap ghost borders

• Input prompt is now inside one block with live slash suggestions directly below it

Features

• Interactive TUI chat and one-shot task mode
• Slash command palette: type / and live suggestions appear
• Live slash-command suggestions appear while typing; /ex then Enter runs /exit, /mo then Enter runs /model
• Slash commands:
  • /help show commands
  • /status show provider/model/session/root
  • /provider switch Anthropic/OpenAI-compatible/NVIDIA NIM provider
  • /apikey enter and save API key for current provider
  • /logout forget saved API key for current provider
  • /model change model; on NVIDIA this fetches the live model list automatically
  • /tools list available agent tools
  • /sessions list saved sessions
  • /resume resume a saved session
  • /save save current session
  • /clear clear conversation context
  • /yes toggle auto-approval
  • /cwd change project root
  • /compact trim conversation context
  • /exit quit
• Anthropic API, OpenAI-compatible APIs, and NVIDIA NIM support
• Tool-calling agent loop
• Safe project-root file tools:
  • list files
  • read files
  • write files
  • patch/replace text
  • grep/search
  • run shell commands
• Approval prompts for writes/shell by default
• --yes automation mode
• Session saving under .ccode/sessions
• Config stored in ~/.ccode/config.json

API keys

Set any provider key ahead of time, or let the CLI ask when needed. When entered in the CLI, keys are saved locally in ~/.ccode/keys.json with file mode 0600. Use /logout to forget the current provider key.

export ANTHROPIC_API_KEY="..."
export OPENAI_API_KEY="..."
export NVIDIA_API_KEY="..."

For OpenAI-compatible providers:

export OPENAI_BASE_URL="https://api.openai.com/v1"
export OPENAI_MODEL="gpt-4o-mini"

For NVIDIA NIM:

export NVIDIA_API_KEY="..."
export NVIDIA_MODEL="nvidia/llama-3.1-nemotron-ultra-253b-v1"

NVIDIA NIM always uses:

https://integrate.api.nvidia.com/v1

For Anthropic:

export ANTHROPIC_MODEL="claude-3-5-sonnet-latest"

Usage

Start the TUI:

ccode

Or explicitly:

ccode chat

One-shot:

ccode run "Add a README and package.json" --yes

Configure defaults:

ccode config set provider anthropic
ccode config set anthropicModel claude-3-5-sonnet-latest
ccode config show

Use NVIDIA NIM:

ccode config set provider nvidia
ccode config set nvidiaModel nvidia/llama-3.1-nemotron-ultra-253b-v1
ccode

Use OpenAI-compatible mode:

ccode config set provider openai
ccode config set openaiModel gpt-4o-mini
ccode run "Explain this project"

Check setup:

ccode doctor

Local development

git clone https://github.com/anmolpromax8-creator/Claude-code.git
cd Claude-code
npm install
npm run build
npm link
ccode --help

Safety

The CLI confines file tools to the current project root. It asks before file writes and shell commands unless --yes is used.