wardayacode

AI-powered coding agent for the terminal. Supports Claude, GPT-4, and Gemini out of the box.

$ wardayacode
❯ Fix the authentication bug in src/auth.ts

● read_file(src/auth.ts)
✓ Found issue: token expiry check uses wrong comparison
● edit_file(src/auth.ts)
✓ Fixed: changed > to < in token expiry check on line 42

The bug was on line 42 — the expiry comparison was inverted.
I've corrected it and the token validation should now work as expected.

Install

npm install -g wardayacode

Requires Node.js 20+.

Quick Start

1. Set your API key

# Anthropic (Claude)
export ANTHROPIC_API_KEY=sk-ant-...

# OpenAI
export OPENAI_API_KEY=sk-...

# Google (Gemini)
export GOOGLE_GENERATIVE_AI_API_KEY=AIza...

Or save it permanently:

wardayacode auth login anthropic
wardayacode auth login openai
wardayacode auth login google

2. Run it

wardayacode        # interactive TUI
wdc                # short alias

3. Start coding

Type any task in natural language. The agent reads your files, makes edits, runs commands, and explains what it did.

Usage

# Interactive TUI (default)
wardayacode

# Send an initial prompt directly
wardayacode "add input validation to src/api/users.ts"

# Choose a model
wardayacode --model gpt-4o --provider openai
wardayacode --model gemini-2.0-flash --provider google
wardayacode --model claude-sonnet-4-20250514 --provider anthropic

# Set permission mode
wardayacode --mode auto        # approve everything automatically
wardayacode --mode plan        # read-only, no file writes

# Set max API retries (default: 3, exponential backoff)
wardayacode --max-retries 5

# Resume a previous session
wardayacode sessions list
wardayacode --resume <sessionId>

# Debug mode (logs tool calls to ~/.wardayacode/logs/)
wardayacode --debug

# Non-interactive (pipe-friendly)
wardayacode --no-tui "summarize the architecture"

Slash Commands

Type / in the TUI to open the command palette, or use these directly:

| Command | Description | |---------|-------------| | /help | Show all available commands | | /clear | Clear the conversation | | /compact | Manually compact context to free up token budget | | /mode <mode> | Switch permission mode | | /model | Show current model | | /session | Show session info | | /tokens | Show real token usage (from provider) | | /undo | Revert the last file change | | /diff | Show uncommitted git changes | | /checkpoint | Create a git stash checkpoint | | /rollback | Restore last checkpoint | | /login <provider> <key> | Save an API key | | /logout <provider> | Remove a saved API key | | /auth | Show provider auth status | | /exit | Exit wardayacode |

Permission Modes

wardayacode asks before making changes. You control how much it can do automatically:

| Mode | File reads | File writes | Bash / Git | Use when | |------|-----------|------------|-----------|----------| | default | ✅ auto | ❓ prompt | ❓ prompt | Daily use | | plan | ✅ auto | ❌ blocked | ❌ blocked | Review-only | | acceptEdits | ✅ auto | ✅ auto | ❓ prompt | Trusted edits | | auto | ✅ auto | ✅ auto | ✅ auto | Scripting / CI | | internal | ✅ auto | ✅ auto | ✅ auto | Fully trusted |

Switch mode mid-session with /mode <name> or choose "Always allow" when prompted.

Tools

The agent has access to these tools:

| Tool | What it does | |------|-------------| | read_file | Read a file with optional line range | | write_file | Create or overwrite a file | | edit_file | Surgical string-replacement edits | | bash | Run shell commands | | git | Run git commands (status, log, diff, add, commit, push, …) | | glob | Find files by pattern (**/*.ts) | | grep | Search file contents with regex | | list_files | List a directory |

Dangerous operations (force push, rm -rf, dd, etc.) are permanently blocked regardless of permission mode.

Configuration

Create .wardayacode.json in your project root for project-level config, or ~/.config/wardayacode/config.json for global defaults:

{
  "provider": "anthropic",
  "model": "claude-sonnet-4-20250514",
  "permissionMode": "default",
  "maxTokens": 8192,
  "temperature": 0,
  "maxRetries": 3,
  "theme": "dark"
}

API keys can also live in config (though environment variables are preferred):

{
  "apiKeys": {
    "anthropic": "sk-ant-...",
    "openai": "sk-...",
    "google": "AIza..."
  }
}

Config priority (highest wins): CLI flags → project .wardayacode.json → user config → defaults.

Sessions

Conversations are saved automatically as JSONL files in .wardayacode/ in your project directory.

wardayacode sessions list          # list sessions
wardayacode sessions delete <id>   # delete a session
wardayacode --resume <id>          # resume a session

Sessions let you continue where you left off across terminal restarts.

Debug & Logs

wardayacode --debug "fix the bug"

With --debug, all tool calls, results, and errors are written to:

~/.wardayacode/logs/YYYY-MM-DD-<sessionId>.log

Each line is structured JSON (ts, level, msg, meta). Tail it in another terminal:

tail -f ~/.wardayacode/logs/$(ls -t ~/.wardayacode/logs | head -1)

You can also set LOG_LEVEL=debug as an environment variable.

Troubleshooting

command not found: wardayacode

npm install -g wardayacode
# if still not found, check npm global bin is in your PATH:
npm config get prefix   # add <prefix>/bin to PATH

API key errors

wardayacode auth list    # check which providers are configured
wardayacode auth login anthropic   # re-enter the key

Agent makes too many changes Use /mode plan to switch to read-only mode mid-session, or start with --mode plan.

Long conversations slow down or lose context wardayacode automatically compacts context when it nears the token limit. You'll see a system message when this happens. Use /clear to start fresh if needed.

Something went wrong and files were changed

/undo          # revert the last file edit
/rollback      # restore to last git checkpoint
/diff          # see what changed

Development

git clone https://github.com/fawwazmw/wardayacode
cd wardayacode
npm install

npm run dev            # run from source (no build)
npm run build          # bundle to dist/
npm run type-check     # TypeScript strict check
npm run lint           # ESLint
npm test               # Vitest watch mode
npm run test:run       # single CI run

License

MIT — Fawwaz Mufid W