rcc

RCC

Overview Use RCC to build and run self-contained automation robots with isolated Python environments and holotree caching. Provide command sequences, config guidance for robot.yaml and conda.yaml, and pointers to references and templates.

Core Workflow

1. Create or clone a robot: rcc robot init --json, rcc robot init -t <template> -d <dir>, or rcc pull <github-url>.
2. Pre-build the environment after init (required): rcc ht vars -r <dir>/robot.yaml.
3. Run tasks: rcc run -r <dir>/robot.yaml --task "<Task>"; use --dev for devTasks; add --silent for clean output.
4. Inspect or execute in the environment: rcc task shell; rcc task script --silent -- <cmd>; rcc ht vars -r robot.yaml --json.
5. Bundle for distribution: rcc robot bundle --robot robot.yaml --output my-robot.py; run with rcc robot run-from-bundle my-robot.py --task <Task>.

Configuration

• Start from assets/templates/robot.yaml and assets/templates/conda.yaml.
• For human-in-the-loop flows, start from assets/templates/hitl-assistant/ (producer/consumer + Assistant UI + custom adapter).
• Prefer uv in conda.yaml for faster installs and pin versions.
• Before updating any uv pin, check the latest release on PyPI and update all templates consistently. As of February 3, 2026, the latest uv is 0.9.28.
• Update assets/templates/conda*.yaml, assets/templates/package.yaml, and any example/reference snippets that show uv so they stay in sync.
• PyPI release page:

https://pypi.org/project/uv/

• Use environmentConfigs with freeze files before conda.yaml for reproducibility.
• Treat output/environment_*_freeze.yaml as runtime artifacts; only copy to the project root when intentionally freezing.
• Use ROBOT_ROOT and ROBOT_ARTIFACTS for path resolution; RCC resolves relative paths from ROBOT_ROOT.
• In robocorp.tasks runtime, prefer get_output_dir() and get_current_task() over direct env-var reads; keep ROBOT_ARTIFACTS/ROBOT_ROOT as fallback for non-task contexts.

Holotree

• List environments: rcc holotree list.
• Delete a space: rcc holotree delete --space <name>.
• Enable shared cache (admin): rcc holotree shared --enable.

Troubleshooting

• Run diagnostics: rcc configure diagnostics.
• Add --debug, --trace, or --timeline when investigating issues.
• Prefer --silent when you only want task output.
• Run scripts/env_check.py --skip-network for a quick environment health check.
• Run scripts/validate_robot.py path/to/robot.yaml to validate robot.yaml and its conda.yaml (requires PyYAML).

Claude Hooks (Optional) Use these only when mirroring Claude Code hook behavior. Scripts live in scripts/hooks/ and may expect CLAUDE_PROJECT_DIR to be set. The supporting RCC task runner configs are in assets/claude-hooks/robot.yaml and assets/claude-hooks/conda.yaml.

References

• references/reference.md: complete CLI reference.
• references/examples.md: recipes and patterns.
• references/installation.md: installation options.
• references/deployment.md: CI/CD, Docker, and remote patterns.
• references/workitems.md: work item APIs and patterns.
• references/actions.md: Actions/MCP framework usage.
• references/robocorp-python.md: Robocorp Python libraries overview.
• references/rpaframework-assistant.md: RPA Framework Assistant (human-in-the-loop UI).
• references/hooks.md: optional Claude Code hook patterns (only if needed).