FAQ

How is LingXiao different from other AI coding tools?

LingXiao is not a chat shell — it’s an orchestration kernel. One sword cleaves the sky, build what you envision. Key differences:

Expert Panel: Leader + Worker multi-role collaboration, not single-agent dialogue
DAG Orchestration: Dependency-aware task graphs with parallel execution
Full State Sync: WebUI/TUI/CLI three-way real-time synchronization
Recoverable: SQLite persistence, recoverable after crashes
Verification Loop: Structured verification + adversarial validation

What Node.js version is required?

Node.js >=24.0.0. Run lingxiao doctor to check your environment.

Which LLMs are supported?

OpenAI (and compatible: DeepSeek, Qwen, Moonshot/Kimi, Gemini, Groq, SiliconFlow, etc.)
Anthropic

Configured via environment variables or ~/.lingxiao/settings.json. See Connect Models.

Where is data stored?

Type	Location
Global config	`~/.lingxiao/settings.json`
Database	SQLite files under `~/.lingxiao/`
Project-level config	`.lingxiao/` directory
Custom Agents (global)	`~/.lingxiao/agents/`
Custom Agents (project)	`.lingxiao/agents/`
Permission configs	`.lingxiao/permissions.*.json`

How to recover a crashed session?

lingxiao list            # List all sessions
lingxiao --session <id>  # Resume a specific session

All session state (messages, task DAG, agent state) is persisted to SQLite and fully recoverable after crashes.

How to configure permission mode?

Four modes:

Mode	Description
`strict`	All write operations require confirmation (default)
`dev`	Development mode, relaxed confirmation
`networked`	Network mode, allows network tools
`yolo`	No confirmation needed (use with caution)

Set via the /mode slash command or tools.permissionMode in ~/.lingxiao/settings.json.

Permissions can persist to four layers: session, project, local, user.

How to access WebUI?

After starting LingXiao, the terminal prints a tokenized URL, e.g.:

http://127.0.0.1:8080/?token=xxxxxxxx

Port info is written to ~/.lingxiao/port. Hardened mode is auto-enabled for non-localhost bindings.

How to use Git Worktree isolation?

# Start with a worktree
lingxiao start --worktree feature-x

# Manage existing worktrees
lingxiao worktree list
lingxiao worktree remove <name>
lingxiao worktree prune

Each task can run in an isolated Git worktree to avoid working directory conflicts.

How to use background mode?

# Background start
lingxiao start --bg --name my-project

# Daemon mode
lingxiao daemon start -p 8080
lingxiao daemon status
lingxiao daemon stop

# View logs
lingxiao daemon logs <name> -f

How to use Eternal autonomous mode?

# Use slash commands in TUI/WebUI
/eternal <goal>       # Set long-term goal
/eternal status       # Check status
/eternal pause        # Pause
/eternal resume       # Resume

Eternal mode cycles at 30-second base interval with exponential backoff up to 960 seconds. Auto-pauses after 8 consecutive failures.

How to use the local LLM Gateway?

Enable in ~/.lingxiao/settings.json:

{
  "llm_gateway": {
    "enabled": true,
    "provider": "openai",
    "model": "gpt-4o",
    "default_rpm": 60,
    "default_tpm": 200000
  }
}

Access via /llm/openai/v1/** and /llm/anthropic/v1/** endpoints with virtual key authentication.

How to create a custom Agent?

# Create a custom Agent
lingxiao agents create my-agent \
  --description "My Agent" \
  --base-role backend \
  --model gpt-4o-mini

# List all Agents
lingxiao agents list

How to install MCP servers?

Use slash commands in TUI/WebUI:

/mcp list                    # List configured servers
/mcp search <query>          # Search marketplace
/mcp install <entry-id>      # Install
/mcp add-stdio <id> <cmd>    # Manually add stdio server
/mcp add-remote <id> <url>   # Manually add HTTP server