Skip to main content
These docs cover the v0.2.1 public beta release. Use loomloom <command> --help as the source of truth for the version installed on your machine.
Choose the platform that owns your account and credential. Both preset platforms are operational:
Both preset platforms support browser login and API Tokens. A compatible custom Server may also be registered with an API Token after doctor verifies it.
No. You can use the CLI directly. Optional integrations are available for Codex, Claude Code, and OpenClaw.
Use Supported Models for a current snapshot, then query the active environment because availability can change by service configuration, permissions, or region.
No. A model can be available in the environment without being compatible with every template or step. Inspect the template schema and the model list for the relevant execution unit.
No. Validation, precheck, quote, schema inspection, uploads, and downloads are preparation steps. Execution creates a hosted run and requires a current estimate followed by explicit confirmation.
No. The CLI runs locally, but confirmed workflows are executed by the configured remote loomloom service. A local Agent Skill is a wrapper and does not copy hidden prompts or server-side execution logic.
CogFoundry does not train CogFoundry or third-party models on Customer Content unless the user or organization explicitly authorizes it. Selected providers may process Customer Content under their own terms and configured controls. See Data Collection and Provider Logging.
Run loomloom doctor, remove secrets from the output, and open an issue in the loomloom GitHub repository.