Agent Cardroom
Enables users to play virtual-chip poker games like Texas Hold'em and Zha Jin Hua against bot opponents within Codex, with interactive UI or text commands.
README
Agent Cardroom
Agent Cardroom is a local Codex plugin for virtual-chip poker against 1–5 built-in bot opponents. It supports Texas Hold'em and Zha Jin Hua. It has no accounts, real-money play, or persistent game history.

Local installation
Clone the source checkout, then install the package for the current user with
the same python3 executable that the plugin mapping invokes:
git clone https://github.com/GZ-L/agent-cardroom.git
cd agent-cardroom
python3 -m pip install --user .
python3 -m agent_cardroom.mcp_stdio --help
The module check is intentionally after installation. Install the plugin from
this checkout in Codex using the repo-local marketplace manifest at
.agents/plugins/marketplace.json, or install the bundle at
plugins/agent-cardroom/. Both mappings run
python3 -m agent_cardroom.mcp_stdio, so they do not depend on a virtual
environment directory being on Codex's PATH.
For development and the full test suite, create an isolated environment:
python3 -m venv .venv
.venv/bin/python -m pip install -e ".[dev]"
Verification and local adapters
The following commands use this checkout's installed virtual environment. Run the HTTP command only when you want to serve the local adapter; it continues running until interrupted.
.venv/bin/python -m pytest tests/test_acceptance.py -v
.venv/bin/python -m pytest -v
.venv/bin/ruff check src tests
git diff --check
.venv/bin/agent-cardroom-http --host 127.0.0.1 --port 8000
These are the project's portable checks. Codex/plugin-directory manifest validation is performed in a Codex environment during the owner-operated publication process; it is not a GitHub-agnostic checkout command.
For a hosted adapter, sessions expire after 60 idle minutes. To remove local state, stop the adapter and start it again; games are not saved permanently.
Playing
Ask the plugin to start a game, for example:
Start a Texas Hold'em game with 3 bot opponents and 1,000 virtual chips.
In an Agent that supports MCP Apps UI, the plugin follows this opening call by showing an interactive table in the current conversation. Click the legal actions and the 25% / 33% / 75% / 133% / double quick-bet options directly; the table displays only virtual Token balances, the current pot, table stakes, and server-approved cards. It does not use a real wallet, deposits, or withdrawals.
If the UI is unavailable, act from the returned state in text:
Call in my Hold'em game.
Or start Zha Jin Hua:
Start a Zha Jin Hua game with 2 bot opponents and 500 virtual chips.
This project is for entertainment with virtual chips and virtual Tokens only. Do not use it for real-money gambling.
Public submission
Public-directory publication is a manual, identity-bound, owner-operated
process, separate from local installation. The authorized owner must create and
push a public repository, host a stable HTTPS MCP endpoint, complete domain
verification, and submit via the Codex directory portal. Tests do not upload
code, create GitHub repositories, alter DNS, or submit to the OpenAI portal.
The local .mcp.json mapping is only for GitHub-download installation; it is
not accepted as the public submission artifact.
See the submission checklist.
The source website, support tracker, privacy policy, and terms are hosted in
this repository: website,
support,
privacy, and terms.
For a public deployment, set MCP_ALLOWED_HOSTS to a comma-separated list of
your deployment hostnames. The Render service at
agent-cardroom.onrender.com is included as the initial deployment hostname;
add a custom domain before switching the public listing to it.
See Privacy, Terms, and MIT License.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。