envi-classic-remote-sensing
This server provides controlled ENVI Classic 5.6 automation for evidence-backed remote-sensing experiments.
Check whether ENVI Classic 5.6 64-bit is installed and safely launchable.
Launch isolated ENVI sessions under an experiment workspace and create run/evidence directories.
Register teacher-provided data, lab guides, and report templates with SHA-256 hashes.
Freeze an ordered stage and screenshot evidence plan.
List ENVI windows and native menu items.
Invoke ENVI menus by full menu path without pixel-coordinate menu clicking.
Run restricted GUI recipes using menus, waits, relative window clicks, text, keys, sleeps, and screenshots; support dry-run validation.
Capture original screenshots as numbered PNG evidence with metadata and SHA-256 in a manifest.
Register generated output files with hashes and metadata.
Finalize and validate the evidence package, detecting missing inputs, missing/duplicate/out-of-order stages, invalid screenshots, and output counts.
Close only MCP-started ENVI sessions, with force termination requiring an explicit flag.
Enforce safety boundaries: no arbitrary PowerShell/CMD/IDL/Python execution, workspace-scoped reads/writes, and read-only use of experiment inputs.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@envi-classic-remote-sensing 用ENVI Classic按老师给的指导书做完实验,保存原始截图,然后生成Word/PDF报告"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
ENVI Classic MCP
能力组成
envi-classicMCP:发现 ENVI 安装、启动隔离会话、调用原生菜单、执行受限窗口内操作、保存截图并计算 SHA-256。envi-classic-lab-automationSkill:从指导书建立并冻结“步骤—截图”计划,调度实验与证据门禁。remote-sensing-reportSkill:以老师模板和已验真的证据包为依据撰写、排版并检查 Word/PDF。
flowchart LR
A["老师数据"] --> D["冻结步骤计划"]
B["实验指导书"] --> D
C["报告模板"] --> D
D --> E["ENVI Classic MCP 实机执行"]
E --> F["原始 PNG + 哈希 + 输出清单"]
F --> G{"证据门禁"}
G -- "完整" --> H["遥感报告 Skill"]
G -- "缺失" --> E
H --> I["DOCX + PDF"]Related MCP server: desktop-touch-mcp
截图真实性约束
报告只能使用 MCP 捕获或用户明确提供的原始截图。
禁止用 Python、Pillow、OpenCV 或其他图像工具裁剪、标注、拼接、增强、重绘截图。
对照展示时,在 Word 中把两个原始图片对象左右或上下排版;不得生成新的合成位图。
同一截图不能冒充多个实验步骤;截图缺失时证据门禁必须失败。
环境要求
Windows 10/11(64 位);
已安装并能正常启动的 ENVI Classic 5.6(64 位);
Python 3.11 或更高版本;
支持本地 stdio MCP 的 Codex 或其他 MCP 客户端;
生成 Word/PDF 时另需可用的文档处理能力,例如 Microsoft Word 或 LibreOffice。
本地安装
git clone https://github.com/xuedai-mayi/envi-classic-remote-sensing.git
Set-Location .\envi-classic-remote-sensing
powershell.exe -ExecutionPolicy Bypass -File .\scripts\setup.ps1 -Dev运行自动测试:
.\.venv\Scripts\python.exe -m pytest运行 MCP 服务:
powershell.exe -ExecutionPolicy Bypass -File .\scripts\run_mcp.ps1.codex-plugin/plugin.json、.mcp.json 和两个 skills/*/agents/openai.yaml 已包含 Codex 插件元数据。将仓库根目录作为本地插件源加载即可同时启用 MCP 与 Skills。
典型输入目录
my-experiment/
├─ 实验指导书.pdf
├─ 实验报告模板.docx
└─ 实验数据/
├─ image.dat
└─ image.hdr推荐提示词:
使用 ENVI Classic 5.6,严格按老师指导书完成这个实验。登记老师的数据、指导书和模板,冻结全部步骤后逐步实机操作并保存原始截图;对照图只在 Word 中左右或上下排版,禁止修改图片。证据完整后生成 Word 和 PDF。
安全边界
只控制由 MCP 自己启动的 ENVI 进程;
不暴露任意 PowerShell、CMD、IDL 或 Python 执行工具;
原生菜单按命令调用,自绘对话框只允许在所属 ENVI 窗口内使用相对坐标;
实验输入只读使用,新结果写入工作区的
.envi-runs/<run-id>/;强制结束会话必须显式传入
force=true;complete=false时不得生成或宣称实验报告已完成。
更完整的设计与证据格式见 docs/architecture.md 和 docs/evidence-contract.md。
期末复习资料
仓库额外收录了“遥感数字图像处理”课程的两份期末资料,见 final-exam-materials。这些文件不参与 MCP 安装或运行。
开发与贡献
请先阅读 CONTRIBUTING.md。严禁向 issue、PR 或提交记录上传课程数据、个人信息、截图、报告成品和 ENVI 安装文件。
许可证与声明
代码以 MIT License 发布。第三方软件与商标声明见 NOTICE.md。
Available Tools
12 toolsenvi_classic_capture_evidence保存一步实验截图证据A
截取指定受控 ENVI 窗口,按顺序保存 PNG,并把阶段、说明、参数、窗口标题、时间和 SHA-256 写入证据清单。
Inputs: caption (required), stage_id (required), session_id (required), parameters, +3 more. Returns: width, height, sha256, caption, +8 more
envi_classic_close_session关闭受控 ENVI 会话ADestructiveIdempotent
只关闭由本服务启动的指定 ENVI 进程。默认先正常关闭;若存在未保存提示则保留进程。force=true 才会强制终止。
Inputs: session_id (required), force, timeout_seconds. Returns: object
envi_classic_finalize_evidence校验并完成报告证据包A
按指导书顺序检查必需阶段、截图哈希、重复项、顺序和输出文件;生成 JSON 与 Markdown 证据包。complete=false 时不得进入报告生成。
Inputs: session_id (required). Returns: message, complete, input_count, output_count, +10 more
envi_classic_health_check检查 ENVI Classic 5.6 环境ARead-onlyIdempotent
读取本机注册表和安装文件,确认 ENVI Classic 5.6(64 位)能否被安全启动。
Inputs: no named parameters. Returns: ok, envi, notes, platform, +2 more
envi_classic_launch启动受控 ENVI Classic 会话B
在指定实验工作区下创建全新的证据目录并启动 ENVI Classic 5.6。只返回本服务可控制的会话 ID。
Inputs: workspace_root (required), experiment_name (required), mode, timeout_seconds. Returns: mode, run_dir, process_id, session_id, +5 more
envi_classic_list_windows列出会话的 ENVI 窗口ARead-onlyIdempotent
列出指定受控 ENVI 进程当前拥有的可见顶层窗口、标题和几何信息。
Inputs: session_id (required). Returns: windows, session_id
envi_classic_register_inputs登记老师提供的实验输入A
在截图前登记老师的数据、实验指导书和报告模板,逐文件计算 SHA-256 并写入证据清单;输入必须位于实验工作区内。
Inputs: data_paths (required), guide_path (required), session_id (required), template_path (required). Returns: inputs, session_id, manifest_path
envi_classic_register_output登记实验输出文件A
登记实验工作区内已经生成的结果文件并计算哈希。不会修改结果文件,也不允许读取工作区外的文件。
Inputs: path (required), label (required), session_id (required), metadata. Returns: path, label, sha256, metadata, +1 more
envi_classic_run_recipe执行受限 ENVI GUI 配方A
顺序执行最多 100 个受限动作:菜单、等待窗口、窗口内相对点击、文本、按键、短等待和证据截图。dry_run 只验证动作,不操作 GUI。
Inputs: actions (required), session_id (required), dry_run. Returns: dry_run, results, session_id, completed_actions
envi_classic_set_stage_plan冻结指导书步骤与截图计划A
把指导书拆出的有序阶段、证据类型、必填参数和预期窗口写入清单并锁定。锁定后不能替换或降低阶段要求。
Inputs: stages (required), session_id (required), require_output. Returns: locked, session_id, stage_count, manifest_path, +2 more
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 12 tools
Most tools have clearly distinct purposes, but 'invoke_menu' and 'run_recipe' both involve menu operations—the former is a single command, the latter a sequence of actions. This slight overlap is minor compared to the overall clarity.
All tools follow a consistent 'envi_classic_' prefix with snake_case naming and action-orientation (e.g., launch, register_inputs, capture_evidence). The only exception is 'health_check', which is a noun but fits the functional pattern.
With 12 tools, the set is well-scoped for a controlled remote sensing lab workflow, covering setup, execution, evidence collection, and finalization without redundancy.
The tools cover the entire lifecycle: health check, launch, input registration, stage planning, GUI interaction, evidence capture, output registration, finalization, and session closure. No obvious gaps exist for the stated purpose.
Maintenance
Related MCP Connectors
Clean PNG/JPEG screenshots via REST or MCP, with goal-driven multi-step navigation.
Turn screenshots into a Word document with numbered figures and a source list.
Automate cloud Chrome—navigate, click, type, screenshot, run code, record screen video
Build, edit and update narrated, interactive training courses from your documents.
Related MCP Servers
- FlicenseNot gradedqualityBmaintenanceDesktop UI automation for AI agents: screenshots, window management, mouse, keyboard, UI Automation tree, OCR. Single Windows x64 binary, no dependencies.3-
- AlicenseAqualityAmaintenanceWindows desktop automation MCP server — screenshot, mouse, keyboard & UI Automation. Lets LLM agents see and control your Windows desktop directly.30726 npm26MIT
- FlicenseBqualityDmaintenanceEnables GUI automation for controlling PIX4Dmatic on Windows through MCP. Supports launching, focusing, capturing screenshots, sending hotkeys, clicking UI elements, opening projects, starting processing, and checking outputs.18-
- AlicenseNot gradedqualityBmaintenanceEnables an AI agent to operate Windows applications through vision-driven UI Automation and record polished demo videos with pre-click camera zoom, narration, and cinematic effects.MIT