FoodforThought CLI

GitHub-like CLI tool for the FoodforThought robotics repository platform.

Version: 0.2.1
PyPI: https://pypi.org/project/foodforthought-cli/

Naming guide (CLI vs SDK vs package)

FoodforThought ships two separate Python tools. They have different names on purpose:

Rule of thumb: use the ate CLI (foodforthought-cli) for the terminal workflow (init, commit, push, generate, deploy); use the kindly SDK to load skill data and check compatibility from Python. See the Kindly SDK README for the SDK.

Installation

pip install foodforthought-cli

Or install from source:

cd foodforthought-cli
pip install -e .

Configuration

Set environment variables:

export ATE_API_URL="https://kindly.fyi/api"
export ATE_API_KEY="your-api-key-here"

Usage

Initialize a repository

ate init my-robot-skill -d "A skill for my robot" -v public

Clone a repository

ate clone <repository-id>

Create a commit

ate commit -m "Add new control algorithm"

Push to remote

ate push -b main

Deploy to robot

ate deploy unitree-r1

Commands

Repository Management

• ate init <name> - Initialize a new repository
• ate clone <repo-id> - Clone a repository
• ate commit -m <message> - Create a commit
• ate push [-b <branch>] - Push commits to remote

Skill Pipeline

• ate pull <skill-id> [--robot <robot>] [--format json|rlds|lerobot] [--output ./data] - Pull skill data for training
• ate upload <video-path> --robot <robot> --task <task> [--project <id>] - Upload demonstrations for labeling
• ate check-transfer --from <source-robot> --to <target-robot> [--skill <id>] - Check skill transfer compatibility
• ate labeling-status <job-id> - Check labeling job status

Parts & Dependencies (New in v0.2.0)

• ate parts list [--category gripper] - List available hardware parts
• ate parts check <skill-id> - Check part compatibility for a skill
• ate parts require <part-id> --skill <skill-id> - Add part dependency to a skill
• ate deps audit - Verify all dependencies are compatible

Text-to-Skill Generation (New in v0.2.0)

• ate generate "<description>" --robot <robot> --output ./skill/ - Generate skill skeleton from natural language

Example:

ate generate "pick up box and place on pallet" --robot franka-panda --output ./new-skill/

Workflow Orchestration (New in v0.2.0)

• ate workflow validate <pipeline.yaml> - Validate a workflow definition
• ate workflow run <pipeline.yaml> --sim - Run workflow in simulation
• ate workflow export <pipeline.yaml> --format ros2 - Export to ROS2 launch format

Team Collaboration (New in v0.2.0)

• ate team create <name> - Create a new team
• ate team invite <email> --role member - Invite a member
• ate team list - List teams and members
• ate skill share <skill-id> --team <team-slug> - Share a skill with a team

Skill Compiler (New in v0.2.1)

Compile skill.yaml specifications into deployable packages. See docs/SKILL_COMPILER.md for full documentation.

# Validate a skill specification
ate validate-skill my_skill.skill.yaml

# Compile to Python package
ate compile my_skill.skill.yaml --target python --output ./dist

# Compile to ROS2 package
ate compile my_skill.skill.yaml --target ros2 --robot robots/ur5.urdf

# Test the compiled skill
ate test-skill ./dist --mode dry-run

# Check robot compatibility
ate check-compatibility my_skill.skill.yaml --robot-urdf robots/ur5.urdf

# Publish to registry
ate publish-skill ./dist --visibility public

Quick Start for AI Assistants:

1. ate_list_primitives - Discover available building blocks
2. ate_validate_skill_spec - Check skill.yaml for errors
3. ate_check_skill_compatibility - Verify robot compatibility
4. ate_compile_skill - Generate deployable code
5. ate_test_compiled_skill - Test without a robot

Dataset Management (New in v0.2.0)

• ate data upload ./sensor-logs/ --skill <id> --stage raw - Upload sensor data
• ate data list --skill <id> --stage annotated - List datasets
• ate data promote <dataset-id> --to skill-abstracted - Promote to next stage
• ate data export <dataset-id> --format rlds - Export dataset

Deployment & Configuration (New in v0.2.0)

• ate deploy --config deploy.yaml --target fleet-alpha - Deploy with config
• ate deploy status <fleet-name> - Check deployment status

Deployment & Testing

• ate deploy <robot-type> - Deploy to a robot (e.g., unitree-r1)
• ate test [-e gazebo|mujoco|pybullet|webots] [-r robot] - Test skills in simulation
• ate benchmark [-t speed|accuracy|robustness|efficiency|all] - Run performance benchmarks
• ate adapt <source-robot> <target-robot> - Adapt skills between robots

Safety & Validation

• ate validate [-c collision|speed|workspace|force|all] - Validate safety and compliance
• ate stream [start|stop|status] [-s sensors...] - Stream sensor data

Cursor IDE Integration (MCP)

FoodforThought CLI includes a full MCP (Model Context Protocol) server for integration with Cursor IDE.

Quick Install via Deep Link

Click this link to install in Cursor:

cursor://anysphere.cursor-deeplink/mcp/install?name=foodforthought&config=eyJtY3BTZXJ2ZXJzIjogeyJmb29kZm9ydGhvdWdodCI6IHsiY29tbWFuZCI6ICJweXRob24iLCAiYXJncyI6IFsiLW0iLCAiYXRlLm1jcF9zZXJ2ZXIiXSwgImVudiI6IHsiQVRFX0FQSV9VUkwiOiAiaHR0cHM6Ly9raW5kbHkuZnlpL2FwaSJ9fX19

Manual Installation

1. Install the CLI:

   pip install foodforthought-cli
   

2. Install MCP SDK:

   pip install mcp
   

3. Add to your Cursor MCP config (~/.cursor/mcp.json or .cursor/mcp.json):

   {
     "mcpServers": {
       "foodforthought": {
         "command": "python",
         "args": ["-m", "ate.mcp_server"],
         "env": {
           "ATE_API_URL": "https://kindly.fyi/api"
         }
       }
     }
   }
   

4. Restart Cursor

Available MCP Tools (37+)

MCP Resources

• skill://{id} - Skill details and configuration
• part://{id} - Hardware part specifications
• workflow://{id} - Workflow definition
• team://{id} - Team information
• deployment://{id} - Deployment status
• audit://{id} - Audit trail

MCP Prompts

• setup_workflow - Guide for multi-skill workflows
• deploy_skill - Step-by-step deployment guide
• debug_compatibility - Debug skill transfer issues
• onboard_robot - Add a new robot to the platform
• audit_deployment - Generate audit reports

See the MCP Server Guide for full documentation.

CI/CD Integration

GitHub Actions

Use the provided actions for CI/CD:

name: Skill CI
on: [push, pull_request]
jobs:
  test:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: kindly-robotics/test-skill@v1
        with:
          skill-path: ./
          robot-model: ur5
      - uses: kindly-robotics/check-transfer@v1
        with:
          skill: .
          min-score: 0.6

License

MIT License. Copyright © 2024-2025 Kindly Robotics. See LICENSE.

The FoodforThought CLI (including the bundled MCP server) is open source under MIT. For Kindly's full open-source posture and per-component licensing, see docs/OPEN_SOURCE.md.