# EJClaw

> Tribunal multi-agent Discord bot framework — autonomous paired review with configurable agent roles

- **URL**: https://www.freshcrate.ai/projects/EJClaw
- **Author**: phj1081
- **Category**: Frameworks
- **Latest version**: `v0.2.3` (2026-04-22)
- **License**: MIT
- **Source**: https://github.com/phj1081/EJClaw
- **Language**: TypeScript
- **GitHub**: 21 stars, 7 forks
- **Registry**: github
- **Tags**: `ai-agent`, `browser-automation`, `bun`, `claude`, `codex`, `discord`, `mixture-of-agents`, `multi-agent`, `typescript`

## Description

Tribunal multi-agent Discord bot framework — autonomous paired review with configurable agent roles

## Recent releases

| Version | Date | Urgency | Changes |
| --- | --- | --- | --- |
| `v0.2.3` | 2026-04-22 | High | ## v0.2.3  ### Added - Added support for room-level reviewer and arbiter agent selection.  ### Changed - Disabled latest owner-final carry-forward by default. - Added lightweight prior-task breadcrumbs so reviewer prompts can retain the previous owner/reviewer conclusions across superseded tasks. - Decoupled `unsafe host mode` from forced fresh Claude reviewer sessions so reviewers resume their session by default.  ### Fixed - Fixed bundled Claude Code binary path resolution to improve runner st |
| `v0.2.1` | 2026-04-13 | High | ## Changed  - Claude / Codex SDK를 최신 patch/minor 수준으로 갱신하고 runner 의존성을 정리했습니다. - reviewer 검증 기준과 continuation/checklist mode 계획 문서를 현재 구조에 맞게 정리했습니다.  ## Fixed  - stale reviewer workspace 레코드가 남아 있을 때 reviewer가 현재 owner worktree 대신 옛 경로를 보는 문제를 수정했습니다. - reviewer 프롬프트를 `EJCLAW_WORK_DIR` 기준 검증으로 고정해 canonical clone을 단독 근거로 오판하는 문제를 완화했습니다. - stale owner run이 buffered progress flush, tracked progress send/edit를 외부로 새는 문제를 수정했습니다. - Claude SDK `task_*` 이벤트가 nested/internal task까지 별도 subagent처럼 표시되던 |
| `v0.2.0` | 2026-04-11 | High | ### Added  - owner / reviewer / arbiter 기반 Tribunal 3-에이전트 흐름 정비 - MoA(Mixture of Agents) 기반 외부 모델 의견 수집 - `assign_room` 기반 room assignment 모델과 `room_settings` SSOT 정착 - 역할별 model / effort / fallback 설정 - reviewer host runtime read-only guard 및 host verification 경로 - Claude OAuth 멀티 토큰 로테이션  ### Changed  - reviewer는 owner workspace를 direct mount로 읽는 현재 모델로 정리 - final delivery 경로를 단일화해 duplicate final 배달 문제를 줄임 - startup precondition을 전용 오류 / exit code 체계로 정리해 crash loop를 막음 - deploy 과정에 `migrat |
| `v0.1.0` | 2026-03-30 | Medium | # EJClaw v0.1.0  First independent release of EJClaw, forked from nanoclaw.  ## Highlights  ### Tribunal 3-Agent System - Owner/Reviewer/Arbiter tri-agent architecture - On-demand arbiter summoning on deadlock detection - Both owner and reviewer can summon arbiter (BLOCKED/NEEDS_CONTEXT) - Mixture of Agents (MoA) — external model opinions from Kimi/GLM  ### Discord-Independent Inter-Agent Communication - `paired_turn_outputs` table for direct agent-to-agent data passing - Discord is display-only |

## Dependency audit

- **Score**: 97/100
- **Total deps**: 23
- **Resolved**: 22
- **Unresolved**: 1
- **License conflicts**: 0
- **Warnings**: 1
- **Scanned**: 2026-04-27

## Citation

- HTML: https://www.freshcrate.ai/projects/EJClaw
- Markdown: https://www.freshcrate.ai/projects/EJClaw.md
- Dependencies JSON: https://www.freshcrate.ai/api/projects/EJClaw/deps

_Generated by freshcrate.ai. Indexes github releases for AI-agent ecosystem packages._