Skip to content
Back to search

agent-council

Multi-agent collaboration plugin for Claude Code - orchestrate multiple AI agents (Codex CLI, Gemini CLI, etc.) for diverse perspectives

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/team-attention--agent-council/badge.svg)](/listings/team-attention--agent-council)

HTML

<a href="/listings/team-attention--agent-council"><img src="/api/listings/team-attention--agent-council/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

Option A: Install via npx (Recommended)

npx github:team-attention/agent-council

This copies the skill files to your current project directory. If you upgrade Agent Council and hit a runtime error like `Missing runtime dependency: yaml`, re-run the installer command above to refresh your installed skill files.

By default, the installer auto-detects whether to install for Claude Code (`.claude/`) and/or Codex CLI (`.codex/`) based on what’s available on your machine and in the repo.

Installed paths: - `.claude/skills/agent-council/` (Claude Code) - `.codex/skills/agent-council/` (Codex CLI)

Optional (Codex repo skill):

npx github:team-attention/agent-council --target codex

Other targets:

npx github:team-attention/agent-council --target claude
npx github:team-attention/agent-council --target both

The generated `council.config.yaml` includes only detected member CLIs (e.g. `claude`, `codex`, `gemini`) and avoids adding the host target as a member. This filtering happens only at initial generation; later edits will not auto-remove missing CLIs.

Option B: Install via Claude Code Plugin (Claude Code only)

# Add the marketplace
/plugin marketplace add team-attention/agent-council

# Install the plugin
/plugin install agent-council@team-attention-plugins

Note (Plugin installs): **Agent Council requires Node.js**, and Claude Code plugins can’t bundle or auto-install Node for you. Install Node separately (e.g. `brew install node` on macOS).

2. Install Agent CLIs

Install the CLIs listed under `council.members` in your `council.config.yaml` (template includes `claude`, `codex`, `gemini`):

# Anthropic Claude Code
# https://claude.ai/code

# OpenAI Codex CLI
# https://github.com/openai/codex

# Google Gemini CLI
# https://github.com/google-gemini/gemini-cli

Verify each member CLI:

command -v claude
command -v codex
command -v gemini

3. Configure Council Members (Optional)

Edit the generated config in your installed skill directory: - `.claude/skills/agent-council/council.config.yaml` - `.codex/skills/agent-council/council.config.yaml`

council:
  chairman:
    role: "auto" # auto|claude|codex|gemini|...
    # command: "codex exec" # optional: run Stage 3 inside council.sh

  members:
    - name: codex
      command: "codex exec"
      emoji: "🤖"
      color: "BLUE"

    - name: gemini
      command: "gemini"
      emoji: "💎"
      color: "GREEN"

    # Add more agents as needed
    # - name: grok
    #   command: "grok"
    #   emoji: "🚀"
    #   color: "MAGENTA"

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 16 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. {
  2. "name": "team-attention-plugins",
  3. "owner": {
  4. "name": "Team Attention"
  5. },
  6. "plugins": [
  7. {
  8. "name": "agent-council",
  9. "source": "./",
  10. "strict": false,
  11. "skills": [
  12. "./skills/agent-council"
  13. ]
  14. }
  15. ]
  16. }
README.md · README · 223 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. # Agent Council
  2. **[한국어 버전 (Korean)](./README.ko.md)**
  3. > A skill that gathers opinions from multiple AI CLIs (Codex, Gemini, ...) and lets a configurable Chairman synthesize a conclusion.
  4. > Inspired by [Karpathy's LLM Council](https://github.com/karpathy/llm-council)
  5. ## Key Difference from LLM Council
  6. **No additional API costs!**
  7. Unlike Karpathy's LLM Council which directly calls each LLM's API (incurring costs), Agent Council uses your installed AI CLIs (Claude Code, Codex CLI, Gemini CLI, ...). This is especially useful if you mainly use one host CLI and occasionally consult others via subscriptions.
  8. Skills are much simpler and more reproducible than MCP. We recommend installing via npx and customizing it yourself!
  9. ## Demo
  10. https://github.com/user-attachments/assets/c550c473-00d2-4def-b7ba-654cc7643e9b
  11. ## How it Works
  12. Agent Council implements a 3-stage process for gathering AI consensus:
  13. **Stage 1: Initial Opinions**
  14. All configured AI agents receive your question simultaneously and respond independently.
  15. **Stage 2: Response Collection**
  16. Responses from each agent are collected and displayed to you in a formatted view.
  17. **Stage 3: Chairman Synthesis**
  18. Your host agent (Claude Code / Codex CLI / etc.) acts as the Chairman by default (`role: auto`), synthesizing all opinions into a final recommendation. Optionally, you can configure a Chairman CLI command to run synthesis inside `council.sh`.
  19. ## Setup
  20. ### Option A: Install via npx (Recommended)
  21. ```bash
  22. npx github:team-attention/agent-council
  23. ```
  24. This copies the skill files to your current project directory.
  25. If you upgrade Agent Council and hit a runtime error like `Missing runtime dependency: yaml`, re-run the installer command above to refresh your installed skill files.
  26. By default, the installer auto-detects whether to install for Claude Code (`.claude/`) and/or Codex CLI (`.codex/`) based on what’s available on your machine and in the repo.
  27. Installed paths:
  28. - `.claude/skills/agent-council/` (Claude Code)
  29. - `.codex/skills/agent-council/` (Codex CLI)
  30. Optional (Codex repo skill):
  31. ```bash
  32. npx github:team-attention/agent-council --target codex
  33. ```
  34. Other targets:
  35. ```bash
  36. npx github:team-attention/agent-council --target claude
  37. npx github:team-attention/agent-council --target both
  38. ```
  39. The generated `council.config.yaml` includes only detected member CLIs (e.g. `claude`, `codex`, `gemini`) and avoids adding the host target as a member. This filtering happens only at initial generation; later edits will not auto-remove missing CLIs.
  40. ### Option B: Install via Claude Code Plugin (Claude Code only)
  41. ```bash
  42. # Add the marketplace
  43. /plugin marketplace add team-attention/agent-council
  44. # Install the plugin
  45. /plugin install agent-council@team-attention-plugins
  46. ```
  47. Note (Plugin installs): **Agent Council requires Node.js**, and Claude Code plugins can’t bundle or auto-install Node for you. Install Node separately (e.g. `brew install node` on macOS).
  48. ### 2. Install Agent CLIs
  49. Install the CLIs listed under `council.members` in your `council.config.yaml` (template includes `claude`, `codex`, `gemini`):
  50. ```bash
  51. # Anthropic Claude Code
  52. # https://claude.ai/code
  53. # OpenAI Codex CLI
  54. # https://github.com/openai/codex
  55. # Google Gemini CLI
  56. # https://github.com/google-gemini/gemini-cli
  57. ```
  58. Verify each member CLI:
  59. ```bash
  60. command -v claude
  61. command -v codex
  62. command -v gemini
  63. ```
  64. ### 3. Configure Council Members (Optional)
  65. Edit the generated config in your installed skill directory:
  66. - `.claude/skills/agent-council/council.config.yaml`
  67. - `.codex/skills/agent-council/council.config.yaml`
  68. ```yaml
  69. council:
  70. chairman:
  71. role: "auto" # auto|claude|codex|gemini|...
  72. # command: "codex exec" # optional: run Stage 3 inside council.sh
  73. members:
  74. - name: codex
  75. command: "codex exec"
  76. emoji: "🤖"
  77. color: "BLUE"
  78. - name: gemini
  79. command: "gemini"
  80. emoji: "💎"
  81. color: "GREEN"
  82. # Add more agents as needed
  83. # - name: grok
  84. # command: "grok"
  85. # emoji: "🚀"
  86. # color: "MAGENTA"
  87. ```
  88. ## Usage
  89. ### Via your host agent (Claude Code / Codex CLI)
  90. Ask your host agent to summon the council:
  91. ```
  92. "Let's hear opinions from other AIs"
  93. "Summon the council"
  94. "Review this from multiple perspectives"
  95. "Ask codex and gemini for their opinions"
  96. ```
  97. ### Direct Script Execution
  98. ```bash
  99. JOB_DIR=$(.codex/skills/agent-council/scripts/council.sh start "Your question here")
  100. .codex/skills/agent-council/scripts/council.sh status --text "$JOB_DIR"
  101. .codex/skills/agent-council/scripts/council.sh results "$JOB_DIR"
  102. .codex/skills/agent-council/scripts/council.sh clean "$JOB_DIR"
  103. ```
  104. Tip: add `--verbose` to `status --text` to include per-member lines.
  105. Tip: use `status --checklist` for a compact checkbox view (handy in Codex/Claude tool cells).
  106. Tip: use `wait` to block until meaningful progress without spamming tool cells (prints JSON, persists a cursor automatically; auto-batches to a small number of updates (typically ~5–10); `--bucket 1` for every completion).
  107. One-shot (runs job → waits → prints results → cleans):
  108. ```bash
  109. .codex/skills/agent-council/scripts/council.sh "Your question here"
  110. ```
  111. Note: In host-agent tool UIs (Codex CLI / Claude Code), one-shot does **not** block. It returns a single `wait` JSON payload so the host agent can update native plan/todo UIs. Continue with `wait` → native UI update → `results` → `clean`.
  112. #### Progress
  113. - In a real terminal, one-shot prints periodic progress lines as members complete.
  114. - In host-agent tool UIs, one-shot returns `wait` JSON (so the host can update native plan/todo UIs).
  115. - Job mode is still available for scripting (`start` → `status` → `results` → `clean`).
  116. ## Example
  117. ```
  118. User: "React vs Vue for a new dashboard project - summon the council"
  119. Host agent (Claude Code / Codex CLI):
  120. 1. Executes council.sh to collect opinions from configured members (e.g., Codex, Gemini)
  121. 2. Displays each agent's perspective
  122. 3. Synthesizes as Chairman:
  123. "Based on the council's input, considering your dashboard's
  124. data visualization needs and team's familiarity, I recommend..."
  125. ```
  126. ## Project Structure
  127. ```
  128. agent-council/
  129. ├── .claude-plugin/
  130. │ └── marketplace.json # Marketplace config (Claude Code only)
  131. ├── bin/
  132. │ └── install.js # npx installer
  133. ├── skills/
  134. │ └── agent-council/
  135. │ ├── SKILL.md # Skill documentation
  136. │ └── scripts/
  137. │ ├── council.sh # Execution script
  138. │ ├── council-job.sh # Background job runner (pollable)
  139. │ ├── council-job.js # Job runner implementation
  140. │ └── council-job-worker.js # Per-member worker
  141. ├── council.config.yaml # Council member configuration
  142. ├── README.md # This file
  143. ├── README.ko.md # Korean documentation
  144. └── LICENSE
  145. ```
  146. ## Notes
  147. - Response time depends on the slowest agent (parallel execution)
  148. - Do not share sensitive information with the council
  149. - Agents run in parallel by default for faster responses
  150. - Subscription plans for each CLI tool are required (no additional API costs)
  151. ## Contributing
  152. Contributions are welcome! Feel free to:
  153. - Add support for new AI agents
  154. - Improve the synthesis process
  155. - Enhance the configuration options
  156. ## License
  157. MIT License - see [LICENSE](./LICENSE) for details.
  158. ## Credits
  159. - Inspired by [Karpathy's LLM Council](https://github.com/karpathy/llm-council)
  160. - Built for [Claude Code](https://claude.ai/code) and [Codex CLI](https://github.com/openai/codex)

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.