loomlc — loom lifecycle.
loomlc is a provider-agnostic orchestrator that runs tasks through a configurable lifecycle of
agent steps, where each step can be executed by a different AI provider's CLI — Anthropic
claude, Google gemini, OpenAI codex, etc. — over pluggable task sources (GitHub, GitLab,
Jira, Linear, local files) and outputs (PRs, MRs, comments, files).
It is not a model harness. It does not call model APIs or reimplement agent loops. It installs and shells out to each provider's own CLI, and orchestrates them: pickup → run steps → verify → emit output, with isolation, scoped credentials, backpressure, and resume.
loomlc generalizes the SDLC agent flow built in the sibling repo strive-ui.io
(plan → engineer ↔ QA → PR, driven by a local runner + .claude/** agents/commands). That flow
proved the pattern; loomlc lifts it into a provider-portable, source-agnostic, generic task-lifecycle
engine — the engineering (SDLC) flow becomes just one shipped preset.
Phase 0 in progress. loomlc is being built in Go, starting with the sdlc lifecycle; only the CLI
scaffold exists so far. See PLAN.md for the design and roadmap and
docs/examples/loomlc.yml for a sample config.
loomlc is licensed under the Apache License 2.0.