控制平面
AI Skill Hub 强烈推荐：控制平面 是一款优质的MCP工具。AI 综合评分 8.0 分，在同类工具中表现稳健。如果你正在寻找可靠的MCP工具解决方案，这是一个值得深入了解的选择。
📚 深度解析
通过安装 控制平面，你的 AI 助手将获得额外的工具调用能力，可以用自然语言直接操控该工具的功能，无需学习复杂的命令行语法。MCP 工具的核心价值在于"一次配置，永久增强"——配置完成后，每次与 AI 对话时都可以无缝调用这些工具。
在技术实现上，MCP 工具通过标准的 JSON-RPC 协议与 AI 客户端通信，工具的功能以"工具列表"的形式暴露给 AI 模型，AI 可以按需调用。控制平面 提供了结构化的工具调用接口，使 AI 模型能够精确地理解和使用每个功能点，显著降低 AI 在工具使用上的错误率。
与传统的 API 集成相比，MCP 工具的优势在于无需编写代码——用户只需在配置文件中添加几行 JSON，即可让 AI 获得全新能力。AI Skill Hub 将 控制平面 评为 AI 评分 8.0 分，属于同类工具中的优质选择。
📋 工具概览
控制平面 是一款遵循 MCP（Model Context Protocol）标准协议的 AI 工具扩展。通过 MCP 协议，它可以让 Claude、Cursor 等主流 AI 客户端直接访问和操作外部工具、数据源和服务，实现 AI 能力的无缝扩展。无论是文件操作、数据库查询还是 API 调用，都可以通过自然语言在 AI 对话中直接触发，极大提升生产效率。
📖 中文文档
控制平面 是一款遵循 MCP（Model Context Protocol）标准协议的 AI 工具扩展。通过 MCP 协议，它可以让 Claude、Cursor 等主流 AI 客户端直接访问和操作外部工具、数据源和服务，实现 AI 能力的无缝扩展。无论是文件操作、数据库查询还是 API 调用，都可以通过自然语言在 AI 对话中直接触发，极大提升生产效率。
- 通过标准 MCP 协议与 Claude、Cursor 等主流 AI 客户端深度集成
- 提供结构化工具调用接口，显著降低 AI 集成复杂度
- 支持 Claude Desktop 和 Claude Code 无缝接入，开箱即用
- 可与其他 MCP 工具组合叠加，构建完整 AI 工作站
- 轻量无侵入设计，不影响现有系统架构
- 在 Claude Desktop 对话中直接调用本地工具，实现 AI 与系统的深度联动
- 通过自然语言驱动复杂的多步骤自动化任务，代替繁琐手动操作
- 将多个 MCP 工具组合使用，构建个人专属 AI 工作站
# 方式一：通过 Claude Code CLI 一键安装
claude skill install https://github.com/aryaminus/controlkeel
# 方式二：手动配置 claude_desktop_config.json
{
"mcpServers": {
"----": {
"command": "npx",
"args": ["-y", "controlkeel"]
}
}
}
# 配置文件位置
# macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
# Windows: %APPDATA%/Claude/claude_desktop_config.json
- 确认已安装 Node.js（v18 或以上版本）
- 打开 Claude Desktop 或 Claude Code 的 MCP 配置文件
- 按「交给 Agent 安装 → Claude Desktop」标签中的 JSON 配置填入 mcpServers 字段
- 保存配置文件并重启 Claude 客户端
- 重启后，在对话中即可使用本工具
# 安装后在 Claude 对话中直接使用 # 示例： 用户: 请帮我用 控制平面 执行以下任务... Claude: [自动调用 控制平面 MCP 工具处理请求] # 查看可用工具列表 # 在 Claude 中输入："列出所有可用的 MCP 工具"
// claude_desktop_config.json 配置示例
{
"mcpServers": {
"____": {
"command": "npx",
"args": ["-y", "controlkeel"],
"env": {
// "API_KEY": "your-api-key-here"
}
}
}
}
// 保存后重启 Claude Desktop 生效
ControlKeel
Agent output is cheap. Governed delivery is not. Keep engineers at the helm.
ControlKeel shapes itself around how you actually steer. It makes your judgment legible, durable, and enforceable - and learns from every decision you make (approved, denied, escalated, dismissed) so it aligns to your real working pattern, not a spec doc you might not re-read. Under that surface, CK is the control plane for agent-led software delivery. It sits between your coding agents and production as a portable "company brain": comparing intended delivery against actual delivery and turning raw agent intent into audited tasks through findings and proofs. It catches the governance drift by enforcing validation and review gates, while keeping work resumable across any host.
---
One-line setup via your agent
Copy/paste this into your agent (OpenCode, Codex, Claude, or another supported host):
Set up ControlKeel for this repository. Read and follow https://raw.githubusercontent.com/aryaminus/controlkeel/main/README.md, https://raw.githubusercontent.com/aryaminus/controlkeel/main/docs/getting-started.md, https://raw.githubusercontent.com/aryaminus/controlkeel/main/docs/one-line-install.md, https://raw.githubusercontent.com/aryaminus/controlkeel/main/docs/direct-host-installs.md, https://raw.githubusercontent.com/aryaminus/controlkeel/main/docs/support-matrix.md, and https://raw.githubusercontent.com/aryaminus/controlkeel/main/docs/agent-integrations.md. Detect this host, install CK if missing, run controlkeel setup, then attach the strongest useful host path with plugin and MCP plus skills/hooks/agents as available. If CK was added only as MCP, call ck_attach for this host to install the missing hooks, skills, commands, and agent profiles. Run controlkeel attach doctor, provider doctor, status, findings, controlkeel me, and the host-specific MCP check; apply only safe local fixes and re-verify. Redact proxy tokens and secrets from shared logs. Pause and ask before continuing if the host needs trust, manual provider config, a restart after attach/plugin changes, or a plan-review approval that cannot auto-wait. For Codex, ensure the project is trusted and restart Codex after attach/plugin changes.Install ControlKeel
```bash
release installers
curl -fsSL https://github.com/aryaminus/controlkeel/releases/latest/download/install.sh | sh
For host-specific one-liners and MCP-only copy/paste setup, see docs/one-line-install.md. If a user starts with MCP-only setup, the agent can call ck_attach to add the repo-local hooks, skills, commands, and companion agents without leaving the session.
Quick start
6. Use guided CLI help whenever you need it
controlkeel help controlkeel help opencode controlkeel help "how do i attach codex" ```
For a full first-run walkthrough, see docs/getting-started.md. For large codebase deployment patterns, see docs/large-codebase-patterns.md.
---
2. In the target repo, bootstrap and inspect the environment
controlkeel setup
How OpenCode is configured with ControlKeel
OpenCode is the primary host used in the benchmark evidence above, and CK supports it through two complementary paths:
controlkeel attach opencodewrites repo-local.opencode/assets, MCP configuration, commands, agents, skills, and.agents/skillscompatibility copies.- The published
@aryaminus/controlkeel-opencodecompanion can be added toopencode.jsonfor the direct plugin-package path. - OpenCode can call
ck_context/ck_context_packto reacquire bounded session state, current task, proof summary, memory hits, resume packet, budget summary, review gate state, and workspace context without relying on chat history. - OpenCode can call
ck_validate,ck_review_submit,ck_memory_record, andck_budgetso validation, approvals, durable memory, and spend evidence stay in CK rather than in one host runtime.
The same governed loop is available to OpenCode, Codex, Claude Code, Copilot, and other supported hosts, but the README guidance leads with OpenCode because that is the best current host-backed evidence path in this repository.
---
Core Bootstrap Package
| Surface | Version | Install / use |
|---|---|---|
| ControlKeel CLI bootstrap | [](https://www.npmjs.com/package/@aryaminus/controlkeel) | npm i -g @aryaminus/controlkeel |
This is the required foundation - install this first before using any other ControlKeel packages or features.
Companion Packages
Published npm packages for direct host integration:
| Package | Host | Version | Install |
|---|---|---|---|
| OpenCode companion | OpenCode | [](https://www.npmjs.com/package/@aryaminus/controlkeel-opencode) | Add "plugin": ["@aryaminus/controlkeel-opencode"] to opencode.json |
| Pi extension | Pi | [](https://www.npmjs.com/package/@aryaminus/controlkeel-pi-extension) | pi install npm:@aryaminus/controlkeel-pi-extension |
Note: After installing companion packages, also run controlkeel attach <host> for the full repo-local experience with commands, agents, and MCP config.
Why use ControlKeel? Benchmark-backed comparison
ControlKeel adds a governance layer around agent output: fast deterministic checks, optional in-agent CK validation, review gates, proof, and budget visibility. The table below is intentionally user-facing: it shows what a team gets from each level of CK integration without requiring you to run the benchmark yourself. Full reproducibility details and caveats live in docs/benchmark-evidence.md.
OpenCode / GPT-5.5 comparison (`host_comparison_v1`, 12 risky scenarios)
| Option | What it means | Catch | Block | Median time | Tokens | Best use |
|---|---|---|---|---|---|---|
| Raw OpenCode | Ask the model and trust the answer | 1/12 | 0/12 | 17,050 ms | 290,327 | Baseline only; not enough for risky changes |
| CK-attached | CK is installed/available, model may call it | 4/12 | 3/12 | **10,818 ms** | 254,581 | Lightweight default when you want CK available without forcing tool use |
| Exhaustive CK-active | Ask the model to inspect every CK surface | 2/12 | 0/12 | 47,560 ms | 510,280 | Demonstrates surface availability, but too slow/expensive for routine use |
| **CK-bounded active** | Model calls CK context + validation, then stops | **5/12** | **3/12** | 23,772 ms | **255,941** | Best practical active-governance tradeoff so far |
| **CK deterministic scanner** | CK validates directly, no model required | **12/12** | **9/12** | **~50 ms** | **0 provider tokens** | Fastest enforcement baseline; ideal for preflight and CI-style checks |
What users should take away:
- Security lift: CK raises systematic detection from raw model output's 1/12 to 5/12 with bounded active governance, and 12/12 with direct deterministic validation.
- Efficiency: bounded active used about half the tokens of exhaustive active while catching more issues.
- Cost control: OpenCode reported
$0cost in JSON events, so we treat tokens/time as the reliable cost proxy. Direct CK scanning uses no provider tokens. - Practical workflow: use deterministic CK validation as the fast gate, and use bounded active governance when you want the agent itself to consult CK before responding.
高质量的MCP工具，值得关注
- 需要 controlkeel 解决具体问题的开发者与运营人员
- 先在测试环境跑通最小用例，再接入生产数据
- API key 直接提交到 git 仓库（请用 .env 并加入 .gitignore）
- 云端托管：可放在 Vercel / Railway / Fly.io 等 PaaS 平台
⚡ 核心功能
- 通过标准 MCP 协议与 Claude、Cursor 等主流 AI 客户端深度集成
- 提供结构化工具调用接口，显著降低 AI 集成复杂度
- 支持 Claude Desktop 和 Claude Code 无缝接入，开箱即用
- 可与其他 MCP 工具组合叠加，构建完整 AI 工作站
- 轻量无侵入设计，不影响现有系统架构
- 需要 controlkeel 解决具体问题的开发者与运营人员
- 先在测试环境跑通最小用例，再接入生产数据
- API key 直接提交到 git 仓库（请用 .env 并加入 .gitignore）
👥 适合人群
🎯 使用场景
- 在 Claude Desktop 对话中直接调用本地工具，实现 AI 与系统的深度联动
- 通过自然语言驱动复杂的多步骤自动化任务，代替繁琐手动操作
- 将多个 MCP 工具组合使用，构建个人专属 AI 工作站
⚖️ 优点与不足
- +标准化 MCP 协议，生态互联性强
- +与 Claude 官方生态无缝对接
- +即插即用，配置简单快捷
- −依赖 Claude 客户端，非 Claude 用户无法使用
- −MCP 协议仍在持续演进，接口可能变更
- −需要一定的配置步骤
该工具使用 NOASSERTION 协议，商用场景请仔细阅读协议条款，必要时咨询法律意见。
AI Skill Hub 为第三方内容聚合平台，本页面信息基于公开数据整理，不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后，再部署至生产环境，并做好必要的安全评估。
📄 NOASSERTION — 请查阅原始协议条款了解具体使用限制。
🔗 相关工具推荐
❓ 常见问题 FAQ
总体来看，控制平面 是一款质量优秀的MCP工具，在同类工具中具备一定竞争力。AI Skill Hub 将持续追踪其更新动态，建议收藏备用，结合自身场景选择合适时机引入使用。
| 原始名称 | controlkeel |
| 原始描述 | 开源MCP工具：Control plane for governed AI coding: validate agent changes, enforce policy, tr。⭐8 · Elixir |
| Topics | ai-governancecode-reviewcompliance |
| GitHub | https://github.com/aryaminus/controlkeel |
| License | NOASSERTION |
| 语言 | Elixir |
收录时间：2026-05-30 · 更新时间：2026-05-30 · License：NOASSERTION · AI Skill Hub 不对第三方内容的准确性作法律背书。