7 个 MCP 工具 · 249 条 TCP 指令 · 在 Cursor、Claude、Codex 中直接开发和控制机械臂 7 MCP tools · 249 TCP commands · Develop and control robot arms directly from Cursor, Claude, and Codex


开发者在 IDE 中说一句话,AI 通过 MCP 获取精确 API 签名,生成工程代码,通过 TCP 直连机械臂执行。出了错,AI 秒级诊断并自动修复。 Developer types one sentence in their IDE. AI calls MCP for exact API signatures, generates project code, sends commands to the robot via TCP. When errors occur, AI diagnoses and fixes in seconds.

AI 搜网页得到的是非结构化文本——可能过时、可能混淆版本、不知道什么操作是危险的。MCP Server 给 AI 的是结构化知识加安全规则,确保生成的代码是对的、安全的。 AI web scraping returns unstructured text — possibly outdated, version-confused, unaware of dangerous operations. MCP Server provides structured knowledge with safety rules, ensuring generated code is correct and safe.
在 IDE 中对话,AI 自动调用 MCP,从写代码到控制真机 Talk in your IDE. AI calls MCP automatically — from writing code to controlling real robots.
force_grasp.py,签名、参数、安全提示全部正确AI combines 3 tool responses, generates complete force_grasp.py with correct signatures and safety notes每个回答:结论 → 代码 → 验证方法 → 如果不行 → 置信度 Every response: conclusion → code → verification → fallback → confidence
添加到 IDE 配置,立即获得机器人开发能力 Add to your IDE config, start building with robot intelligence
control_robot 会向机械臂发送真实 TCP 指令。已通过 6 轮对抗性测试 + 真机验证。
问题反馈:GitHub Issuescontrol_robot sends real TCP commands to robot arms. Passed 6 rounds of adversarial testing + real robot verification.
Report issues: GitHub Issues