By HotRedMat
Multi-agent orchestrator plugin for Claude Code — agents perform in concert through a 5-phase pipeline (Gather → Deliberate → Execute → Audit → Document) with Reuse-First cross-cutting policy.
Ensembra 의 아키텍처 설계 담당. 모듈 경계·구조 패턴·설계 결정을 다룬다. Phase 1/3 참여. 기본 Transport 는 MCP(gemini-ensembra) 이며, 폴백 순서는 Ollama(기본 qwen2.5:14b — v0.10.0+ config 로 변경 가능) → Claude(sonnet). 신규 기능·리팩토링·구조 변경 토론 시 호출한다.
Ensembra 의 구현 전략 담당. 패턴·라이브러리·API 선택, 기존 코드 스타일 준수, 실제 실행 가능한 Plan 작성을 책임진다. Phase 1/3 참여. Phase 2 실제 실행자(Claude Code 본체)와 같은 엔진 계열이라 Plan→실행 간극 최소화. 구현 레벨 의사결정 토론 시 호출한다.
Ensembra 의 반론 전담. 숨은 가정 드러내기, 과잉 설계 경고, "고치지 말아야 할 이유" 제시를 책임진다. Phase 1/3 참여. 모델 haiku 로 빠른 반응성. 반론이 필요한 모든 토론 라운드에서 호출한다.
Ensembra v0.8.0+ 의 최종 감사자. Phase 3 에서 모든 preset 공통으로 전문 감사자 검토 뒤에 배치되어 만장일치 판정을 내린다. Claude Code opus 전용. Phase 1 토론에는 참여하지 않는다. 토론 합의 + final-auditor verdict 가 모두 충족될 때만 "unanimous" 간주, Phase 4 문서화 단계로 진행한다.
Ensembra 파이프라인의 지휘자(Conductor). 1인 개발자의 요청을 받아 6명의 Performer 서브에이전트(planner/architect/developer/security/qa/devils-advocate)를 Phase 0~4 파이프라인으로 조율한다. Reuse-First 교차 원칙을 강제하고, Phase 2 실행은 Claude Code 본체에 위임하며, Phase 4 문서화는 scribe 에게 위임한다.
Ensembra 의 모든 설정을 선택형 대화로 관리하는 picker. 모델·프리셋·라운드·Deep Scan·Transport·타임아웃·로깅·Reports·Reuse-First 를 통합 관리한다. 사용법 "/ensembra:config". Claude Code /config 와 유사한 UX.
Ensembra 일일/주간 업무 보고서 생성. 사용법 "/ensembra:report daily" 또는 "/ensembra:report weekly". 자동 생성 아님, 사용자 명시 호출만.
Ensembra 파이프라인의 메인 진입점. 프리셋과 요청을 받아 Phase 0~4 를 순차 실행한다. 사용법 "/ensembra:run <preset> <요청>" 예 "/ensembra:run feature 결제에 쿠폰 할인 추가". 프리셋 feature/bugfix/refactor/ops/ops-safe/security-audit/source-analysis 중 선택 (v0.9.0+ ops/ops-safe 신설).
Ensembra 인수인계서 생성 스킬. 프로젝트를 다른 담당자에게 넘기기 위한 독립 문서를 만든다. 사용법 "/ensembra:transfer" (전체) 또는 "/ensembra:transfer <경로 또는 자연어>". 세션 중단용 handoff 와 다르며 handoff 는 외부 플러그인(d2-ops-handoff) 담당.
Admin access level
Server config contains admin-level keywords
Requires secrets
Needs API keys or credentials to function
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
This plugin requires configuration values that are prompted when the plugin is enabled. Sensitive values are stored in your system keychain.
gemini_api_keyGoogle AI Studio API key. Stored in OS keychain (macOS Keychain / Linux Secret Service / Windows Credential Manager) via Claude Code's sensitive config channel. Passed to the MCP server process env — never exposed in skill/agent content. Set via /plugin → ensembra → Configure options. If not set, architect Performer falls back to Ollama → Claude sub-agent.
${user_config.gemini_api_key}ollama_endpointOllama HTTP endpoint for the security, qa, and (opt-in) developer Performers. Default: http://localhost:11434. Leave blank to skip Ollama — those Performers will fall back to Claude sub-agents.
${user_config.ollama_endpoint}developer_transportOpt-in override for the developer Performer transport chain. Values: 'claude' (default, keep Claude sub-agent sonnet — safest for Phase 2 style match), 'external' (use MCP(gemini-2.5-pro) → Ollama(gpt-oss:20b) → Claude sonnet fallback chain — maximizes external LLM use, reduces Claude token cost). Leave blank to use the default. See CONTRACT.md §8.8 Transport Fallback Chain Protocol.
${user_config.developer_transport}
Where agents perform in concert — a multi-agent orchestrator plugin for Claude Code.
![]() /ensembra:run feature — 5-phase pipeline with consensus and reuse evaluation
|
![]() /ensembra:config — interactive picker with cascade-safe custom mode
|
![]() /ensembra:transfer — 10-section handover document with devils-advocate pitfalls
| |
Ensembra is a Claude Code plugin that orchestrates six specialist agents and one scribe through a 5-phase pipeline to produce structured code reviews, mutual supervision, automatic documentation, and project handover documents. Built for solo developers who want team-level deliberation without the team.
Key ideas:
Phase 0 Gather — Deep Scan produces a shared context snapshot
Phase 1 Deliberate — R1 → (optional R2) → Synthesis with peer signatures
Phase 2 Execute — Claude Code edits files per the agreed plan
Phase 3 Audit — designated performers verify the diff
Phase 4 Document — scribe records Task / Design / Request / Daily / Weekly
v0.8.0 splits Performers into three tiers. Opus is forbidden in debate and used only by the final-auditor in Phase 3 — the unanimous-consensus judge.
Debate (Phase 1) — external LLMs + sonnet-or-lower, no opus:
| Role | Responsibility | Default Transport | Default Model |
|---|---|---|---|
| 🧭 planner | Requirements, acceptance criteria | Claude sub-agent | sonnet (v0.8.0: was opus) |
| 🏛 architect | Module boundaries, patterns | MCP (Gemini) → Ollama → Claude | gemini-2.5-flash |
| 🛠 developer | Implementation strategy | Claude sub-agent (opt-in external chain) | sonnet |
| 🛡 security | Threats, secrets, OWASP | Ollama → Claude | qwen2.5:14b |
| 🧪 qa | Edge cases, regression | Ollama → Claude | qwen2.5:14b |
| 😈 devils-advocate | Counter-arguments, YAGNI | Claude sub-agent | haiku |
Audit (Phase 3) — specialist auditors → final-auditor:
| Role | Responsibility | Default Transport | Default Model |
|---|---|---|---|
| (specialists) | Preset-specific specialists (see presets table) | varies | varies |
| ⚖️ final-auditor | Unanimous-consensus judge, always last (v0.8.0) | Claude sub-agent | opus |
Document (Phase 4):
| Role | Responsibility | Default Transport | Default Model |
|---|---|---|---|
| ✍️ scribe | Phase 4 documentation | Claude sub-agent | sonnet |
Unanimous consensus (v0.8.0) = Phase 1 agreement ≥ 70% AND final-auditor.verdict == pass. Final-auditor rework is capped at 1 cycle (opus cost control). See CONTRACT.md §8.8 for the generalized Transport Fallback Chain Protocol and §11.3 for Final Audit details.
All models auto-fall back to Claude sub-agents when the external transport is unavailable.
/ensembra:run <preset> <request> — main pipeline entry point/ensembra:config — unified interactive settings picker (all options, all cascade-safe)/ensembra:transfer [scope] — project handover document (full project, path, or natural-language scope)/ensembra:report daily|weekly — roll-up reportsKarpathy-style LLM-friendly wiki system for Claude Code. Auto-accumulates knowledge across sessions, optimized for low-spec local LLMs.
npx claudepluginhub hotredmat/ensembra --plugin ensembrav9.44.1 — Patch release for Gemini environment/version detection and qwen auth gating. Run /octo:setup.
Multi-agent orchestration via li o flow and li o fanout
Verification-first engineering toolkit for Claude Code. 15 skills across a 5-phase spine (Investigate → Design → Implement → Verify → Ship), 8 specialist agents, an interactive setup wizard. Every skill has rationalizations + evidence requirements. Built for senior ICs and tech leads.
Multi-agent orchestration framework for Claude Code, Gemini CLI, and Codex CLI — 19 agents, 14 skills, 16 commands, quality gates, TDD enforcement
Adversarial multi-agent pipeline for Claude Code. GAN-style loops where generators produce artifacts, discriminators validate them, and feedback drives convergence.
AI development loop — orchestrator distributes tasks to headless workers, independent auditor verifies, structural enforcement auto-blocks downstream on upstream failure. Full-cycle validated with 10-scenario test harness.