The idea: instead of asking one AI coding agent (Claude, Gemini, Codex, Aider, etc.), you group them into your "Agents Parliament."
When you submit a task:
- Your main agent receives the request and decides if it needs help
- It calls other agents as tools — Claude might ask Gemini for a code review, or delegate a refactor to Aider
- The agents collaborate — each bringing their strengths to produce better results
Each agent can also spawn sub-agents on the fly, dynamically creating their instructions based on the task.
| Server | CLI Tool | Publisher |
|---|---|---|
claude-mcp |
Claude Code | Anthropic |
aider-mcp |
Aider | Open Source |
codex-mcp |
OpenAI Codex | OpenAI |
gemini-mcp |
Gemini CLI | |
goose-mcp |
Goose | Block (Square) |
- Python 3.10+
- The respective CLI tools installed (see below)
# Claude Code (required for claude-mcp)
# Install from: https://docs.anthropic.com/en/docs/claude-code
# Aider (required for aider-mcp)
pip install aider-chat
# OpenAI Codex (required for codex-mcp)
npm install -g @openai/codex
# Gemini CLI (required for gemini-mcp)
npm install -g @google/gemini-cli
# Goose (required for goose-mcp)
curl -fsSL https://github.com/block/goose/releases/download/stable/download_cli.sh | bash# Install from PyPI (recommended)
pip install agenters
# OR install from source
git clone https://github.com/roeiba/agenters.git
cd agenters
pip install .The agenters CLI provides easy provisioning for your environment.
See what's currently configured:
agenters status# Provision all agents globally for Cursor
agenters provision --global --client cursor --agents all --yes
# Provision specific agents for a project
agenters provision --project --agents claude,aider
# Preview changes without applying (dry-run)
agenters provision --global --client vscode --dry-run| Client Flag | Description |
|---|---|
--client claude |
Claude Desktop |
--client cursor |
Cursor IDE |
--client vscode |
VS Code |
--client windsurf |
Windsurf |
--client antigravity |
Antigravity (Gemini) |
For an interactive experience:
# Interactive installer with step-by-step prompts
agenters install
# Interactive uninstaller
agenters uninstallOnce installed, you can use the CLI commands directly:
# Claude MCP Server
claude-mcp
# Aider MCP Server
aider-mcp
# Codex MCP Server
codex-mcp
# Gemini MCP Server
gemini-mcp
# Goose MCP Server
goose-mcpAdd to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"claude-agent": {
"command": "claude-mcp",
"args": []
},
"aider-agent": {
"command": "aider-mcp",
"args": []
},
"codex-agent": {
"command": "codex-mcp",
"args": []
},
"gemini-agent": {
"command": "gemini-mcp",
"args": []
},
"goose-agent": {
"command": "goose-mcp",
"args": []
}
}
}| Tool | Description |
|---|---|
ask_claude |
Simple prompt to Claude |
ask_claude_with_system |
Prompt with custom system prompt |
ask_claude_json |
Get structured JSON response |
ask_claude_in_directory |
Run with directory context |
ask_claude_with_tools |
Run with specific tools enabled |
ask_claude_with_hooks |
Execute with pre/post hooks |
ask_claude_with_skill |
Load a skill for specialized tasks |
spawn_claude_agent |
Create a sub-agent with custom instructions |
get_claude_capabilities |
Get agent card for A2A discovery |
get_claude_version |
Get CLI version |
| Tool | Description |
|---|---|
aider_chat |
Send a message to make code changes |
aider_architect |
Use architect mode for planning |
aider_ask |
Ask questions without making changes |
get_aider_capabilities |
Get agent card for A2A discovery |
get_aider_version |
Get Aider version |
| Tool | Description |
|---|---|
codex_prompt |
Send a prompt (suggest mode) |
codex_full_auto |
Run in full-auto sandboxed mode |
codex_auto_edit |
Run in auto-edit mode |
get_codex_capabilities |
Get agent card for A2A discovery |
get_codex_version |
Get Codex CLI version |
| Tool | Description |
|---|---|
gemini_prompt |
Send a prompt to Gemini |
gemini_in_directory |
Run with directory context |
gemini_with_search |
Run with Google Search grounding |
gemini_with_playbook |
Execute with a specialized playbook |
get_gemini_capabilities |
Get agent card for A2A discovery |
get_gemini_version |
Get Gemini CLI version |
| Tool | Description |
|---|---|
goose_run |
Run with text instructions |
goose_run_file |
Run with instructions from file |
goose_run_recipe |
Run a predefined recipe |
goose_with_toolkits |
Run with specific toolkits enabled |
get_goose_capabilities |
Get agent card for A2A discovery |
get_goose_version |
Get Goose version |
The Agents Parliament now supports agent-to-agent (A2A) communication and full mesh coordination.
Each agent exposes a get_*_capabilities tool that returns an agent card:
from agenters.a2a_protocol import discover_agents, find_best_agent
# Discover all agents
agents = discover_agents()
for agent in agents:
print(f"{agent.name}: {agent.strengths}")
# Find best agent for a task
agent = find_best_agent("search the web for latest Python trends")
# Returns: gemini-agent (has search-grounding strength)Route tasks to the best agent based on capabilities:
from agenters.mesh_coordinator import create_mesh
mesh = create_mesh()
# Find best agent for a task
agent, reasoning = mesh.route_to_best_agent("Refactor this code with Git commits")
# agent: aider-agent, reasoning: git-integration strength
# Suggest a team for complex tasks
team = mesh.suggest_team("Research latest APIs, implement feature, and deploy")
# Returns: [(gemini-agent, "Research"), (claude-agent, "Implement"), (goose-agent, "Deploy")]| Feature | Agent | Description |
|---|---|---|
| Hooks | Claude | Execute shell commands before/after prompts |
| Skills | Claude | Load specialized SKILL.md files |
| Sub-agents | Claude | Spawn task-specific agents dynamically |
| Playbooks | Gemini | Pre-configured workflow guides |
| Toolkits | Goose | Enable MCP extensions (GitHub, Jira, etc.) |
| Recipes | Goose | YAML-defined reusable workflows |
| Architect Mode | Aider | Two-model planning approach |
| Full-Auto | Codex | Sandboxed autonomous execution |
For detailed documentation, see docs/advanced_features.md.
MIT