TuhuCar CLI is a production-oriented command-line client for Tuhu car-care knowledge workflows. It provides a consistent interface for querying maintenance guidance, exposing machine-readable schemas for LLM tooling, and installing ready-to-use skills into popular AI coding assistants.

• Purpose-built for car-care knowledge queries instead of generic chat wrappers
• Stable CLI surface with both markdown and JSON envelope output modes
• MCP-based upstream integration with schema introspection for agent workflows
• Built-in skill installation for Claude Code, Cursor, Codex, OpenCode, and Gemini CLI
• Multi-platform distribution via npm, release binaries, and shell installer

• knowledge query for maintenance and ownership questions
• knowledge schema for LLM introspection and tool wiring
• config init and config show for local runtime configuration
• skill install and skill uninstall for assistant integration
• --dry-run support for debugging and prompt/tool development
• Update notices for installed distributions

npm install -g @tuhucar/cli

curl -fsSLO https://raw.githubusercontent.com/tuhucar/cli/main/scripts/install.sh
less install.sh
sh install.sh

brew install tuhucar/tap/tuhucar

# 1. Create default config
tuhucar config init

# 2. Ask a maintenance question
tuhucar knowledge query "2024款大众朗逸1.5L 全合成机油多久换一次？"

# 3. Continue a multi-turn session
tuhucar knowledge query --session-id <session_id> "那刹车油多久换一次？"

# 4. Inspect the machine-readable command schema
tuhucar knowledge schema

Configuration is stored at ~/.tuhucar/config.toml. If you need to point the CLI at another gateway temporarily, set TUHUCAR_ENDPOINT in the environment.

# Human-friendly output
tuhucar knowledge query "机油多久换一次"

# JSON envelope output
tuhucar --format json knowledge query "机油多久换一次"

# Multi-turn dialog
tuhucar knowledge query --session-id <session_id> "轮胎气压怎么判断？"

# Tool schema for agents
tuhucar knowledge schema

tuhucar config init
tuhucar config show

tuhucar skill install
tuhucar skill uninstall

TuhuCar CLI ships with embedded skill assets and platform-specific installation flows.

After installing the CLI, run tuhucar skill install to register the bundled skills on detected platforms.

The ClawHub-ready publish source lives in skills/clawhub/tuhucar-knowledge-assistant/. It is intentionally self-contained and separate from the bundled local assistant skills used by tuhucar skill install.

Example publish flow:

clawhub login

clawhub publish ./skills/clawhub/tuhucar-knowledge-assistant \
  --slug tuhucar-knowledge-assistant \
  --name "TuhuCar Knowledge Assistant" \
  --version 0.0.3 \
  --tags latest \
  --changelog "Address ClawHub security scan findings"

If you update the CLI behavior that the skill depends on, keep these files aligned:

• skills/tuhucar-* for bundled local assistant installs
• skills/clawhub/tuhucar-knowledge-assistant/ for ClawHub publishing

The project is organized as a small Rust workspace with clear module boundaries.

• Rust stable
• Node.js 14+ for npm packaging workflows

# Build the workspace
cargo build --workspace

# Run tests
cargo test --workspace

# Run clippy with CI-level strictness
cargo clippy --workspace --all-targets -- -D warnings

# Verify formatting
cargo fmt --all -- --check

# Run the CLI locally
cargo run -p tuhucar-cli -- --help

• GitHub Releases publish platform binaries
• npm publishes @tuhucar/cli
• The shell installer downloads release artifacts directly from GitHub Releases

Issues and pull requests are welcome. When changing CLI behavior, keep the README examples, integration tests, and schema-facing output aligned.

MIT. See LICENSE.