Ccswarm

ccswarm
Turn uncertain coding work into reviewable changes through Sangha.

ccswarm is a workflow engine for AI coding agents, with Sangha as its product
core: shared acceptance criteria, separate assessments, evidence-backed
objections, revision, and a recorded decision. Claude Code and Codex execute
the work; ccswarm governs how the workflow advances.

The current default flow runs plan → Sangha quorum → implement → review → fix.
Sangha counts approval markers only from completed reviews and rejects an
unreachable quorum. Evidence-bound acceptance, objection
resolution, and checkpoint recovery are the next design, not shipped
guarantees. See Sangha Product Core for the
principles, implementation gaps, and feature priorities.
Astra Integration adds the design for live guidance,
pending tool results, reasoning choices, and bounded delegation, distinguishing
model selection from future backend integration.

Hire ccswarm when

- You want a repeatable plan, assessment, implementation, and review process
  without rebuilding the workflow for each task.
- You want reviewer concerns and verification evidence to inform the decision
  to accept a change; this is the direction of the Sangha core.
- You need recorded runs and timeline comparisons. Replay re-executes the task;
  undo currently reports commits rather than rolling them back.
- You use multiple provider CLIs (Claude Code / Codex, with gh copilot probed for
  diagnostics) and don't want to pick one.

Quick start

doctor reports missing ANTHROPIC_API_KEY as a warning, not a hard failure,
when you use provider CLIs that are already authenticated locally.

scaffold creates a new git repository, writes a minimal npm project with a
passing npm test script, then runs the selected flow in that project. The
global --provider…