GeoMCP
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., "@GeoMCPlist my active jobs and show the status of the latest healthcheck"
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.
GeoMCP
GeoMCP 是面向地球物理科研服务器的安全工具与知识服务平台,让 Codex / AI Agent、人工 CLI 和 Python 脚本共享同一套权限、任务和科学计算接口。
当前状态:v0.1(Step 01–10)
已实现:
Permission / Path Sandbox
Workspace / Data Region I/O
Python API + CLI + MCP
Job Manager (SQLite + JSON mirror)
Local CPU Executor
Fixed-endpoint Remote GPU Executor
GPU Worker Registry
DAS Basic (DASPy)当前 MCP 工具:
system.status
filesystem.inspect
workspace.list
job.list
job.status
job.result
job.cancel
job.submit_healthcheck
das.inspect
das.read_window
das.bandpass
das.rms
das.plotRelated MCP server: SSH Vault MCP
v0.1 封板状态
Step 10.1 Hardening 已补齐:CI 全量 DAS 依赖、SQLite 原子 Job 状态转换、dispatch watchdog、MCP 固定配置边界、CPU/GPU healthcheck 闭环、DASPy 真正按 channel/sample 窗口读取,以及 MCP DAS 窗口结果压缩。
在此基础上加入 Workspace / Data Region:Agent 可以选择管理员配置的科研区域并使用相对输入/输出路径,但 Workspace 不能扩大全局 Path Sandbox。
MCP 不接受 config_dir、host、port、username、command 或任意 task type。
安装
基础 CLI / Python API:
python -m pip install -e .MCP:
python -m pip install -e ".[mcp]"DASPy 支持:
python -m pip install -e ".[das]"开发与完整 v0.1:
python -m pip install -e ".[test,mcp,das]"
pytest -q安全边界
/cluster/datapool2/xuxy/**默认可读仅 GeoMCP
runtime/outputs/knowledge默认可写原始科研数据默认只读
Agent 可通过 Workspace 使用相对输入/输出路径,但 Workspace 不能扩大全局 Path Sandbox
不暴露 delete、任意 Shell、任意 SSH
GPU endpoint 只能由服务器配置指定,Agent 不能传 host/port/user/command
Worker 只执行 Registry 白名单任务
DAS 大窗口和超过 Nyquist 的滤波会被拒绝
1012 作为控制侧;GPU Executor 默认关闭,配置固定 endpoint 后由 1012 把 Job ID 转发给 1015 Worker。
文档
This server cannot be deployed
Maintenance
Related MCP Connectors
Scoped, audited SSH exec, sessions, and SFTP on your saved servers without exposing credentials
Operate Linux, macOS and Windows from your LLM. Every action runs through an auditable allowlist.
Task management for people and AI agents, with scoped OAuth access to issues, projects, and docs.
Task management for people and AI agents, with scoped OAuth access to issues, projects, and docs.
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceEnables AI assistants to manage SLURM cluster jobs with safety guardrails, including file transfer, job submission, log reading, and remote command execution.1MIT
- FlicenseNot gradedqualityDmaintenanceEnables AI agents to securely access encrypted secrets (SSH keys, API tokens, passwords) with real-time user approval via Passkey, and supports SSH remote execution through the MCP protocol.-
- AlicenseNot gradedqualityBmaintenanceBridges AI assistants to seismology software (ObsPy, CWP/SU, SAC, PyGMT), allowing users to process seismic waveforms and perform analysis through natural language instead of writing code.1MIT
- AlicenseNot gradedqualityCmaintenanceEnables AI agents to manage remote servers via SSH, including command execution, multi-host batch operations, SFTP file transfer, background job handling, DevOps diagnostics, port tunneling, and safety guardrails like high-risk command blocking and read-only mode.MIT