convez.ai
MCP ServerIntegração & APIsAvançado
huangzesen

SPEDAS MCP — Servidor de Dados Heliosféricos e Efemérides

huangzesen/spedas-mcp

Acesso integrado a dados heliosféricos, planetários e efemérides SPICE via MCP

Ver no GitHub
InicialPluginScore · de 0 a 100

Servidor MCP unificado que integra dados de observatórios solares (CDAWeb), missões planetárias (PDS) e cálculos de efemérides (SPICE). Permite descoberta de dados, busca de parâmetros e download de séries temporais para análise de fenômenos heliosféricos e órbitas espaciais. Essencial para pesquisadores e engenheiros trabalhando com missões espaciais.

Instalação

$npx convez-ai huangzesen/spedas-mcp

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
1
Acompanhando
0
Saúde
42%
Último commit
há 4 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Análise de dados de observatórios solares (CDAWeb) para pesquisa heliosférica
  • Cálculo de efemérides e distâncias entre corpos celestes e espaçonaves
  • Transformações de coordenadas espaciais e integração com ferramentas de dinâmica orbital
  • Pipelines automatizados de descoberta e download de dados de missões planetárias (PDS)

Veredito

Ferramenta especializada e bem arquitetada para domínio científico específico (astrofísica/heliofísica). Oferece contrato claro entre agente e servidor, mas requer contexto técnico profundo em missões espaciais e sistemas de coordenadas astronômicas. Ideal para pesquisadores e equipes que trabalham regularmente com dados NASA/SPICE.

README

A tradução para o português deste README ainda está na fila.

spedas-mcp

Unified SPEDAS-oriented MCP server that composes focused XHelio building blocks:

  • xhelio-cdaweb: CDAWeb observatory/dataset discovery, parameter metadata, and CDF data fetch.
  • xhelio-pds: NASA PDS Planetary Plasma Interactions mission/dataset discovery, parameter metadata, and PDS data fetch.
  • xhelio-spice: SPICE kernel management, spacecraft/body ephemeris, distances, and coordinate transforms.

The goal is not to replace SPEDAS/PySPEDAS in one step. The first repo boundary is a reliable MCP layer for Claude Code, Codex, and future SPEDAS plugins.

Install

pip install spedas-mcp[mcp]

For development from source:

pip install -e '.[dev,mcp]'
pytest -q

For a CI-safe MCP check that does not fetch CDAWeb data or download SPICE kernels:

uv run --extra mcp python scripts/smoke_mcp_list_tools.py --json

The smoke starts the stdio server with isolated temporary caches, runs MCP initialize + list_tools, and verifies the advertised tool names.

Run

spedas-mcp
python -m spedas_mcp

Optional cache overrides:

spedas-mcp \
  --cdaweb-cache-dir /tmp/spedas-cdaweb-cache \
  --pds-cache-dir /tmp/spedas-pds-cache \
  --spice-kernel-dir /tmp/spedas-spice-kernels

MCP tools

Overview

  • spedas_overview() — summarize available capability groups and workflow.

CDAWeb tools

  • browse_observatories()
  • load_observatory(observatory_id)
  • browse_parameters(dataset_id, dataset_ids?)
  • fetch_data(dataset_id, parameters, start, stop, output_dir, format="csv")
  • manage_cdaweb_cache(action, ...)

PDS PPI tools

  • browse_pds_missions(query?)
  • load_pds_mission(mission_id)
  • browse_pds_parameters(dataset_id?, dataset_ids?)
  • fetch_pds_data(dataset_id, parameters, start, stop, output_dir, format="csv")
  • manage_pds_cache(action, ...)

SPICE tools

  • list_spice_missions()
  • get_ephemeris(target, time, frame="ECLIPJ2000", observer="SUN", output_file="", time_end="", step="1h")
  • compute_distance(target1, target2, time_start, time_end, step="1h")
  • transform_coordinates(vector, time, from_frame, to_frame, spacecraft?)
  • list_coordinate_frames()
  • manage_spice_kernels(action, ...)

Agent workflow contract

  1. Discover before fetching: browse_observatories, load_observatory, browse_parameters, browse_pds_missions, load_pds_mission, browse_pds_parameters, or list_spice_missions first.
  2. Bulk data must go to files (output_dir / output_file), not inline chat.
  3. Use compact summaries: paths, stats, units, coordinate frames, cache size, warnings.
  4. Treat CDAWeb/PDS downloads and SPICE kernel downloads as integration actions; make time ranges small by default.

Agent plugin wrappers

This repo also packages the shared MCP server for coding agents:

  • plugins/spedas-claude/ — Claude Code plugin named spedas-claude, with a Claude plugin manifest, MCP server config, SPEDAS workflow skill, and slash-command prompts.
  • .agents/plugins/spedas-codex/ — Codex plugin named spedas-codex, with a Codex plugin manifest, MCP server config, and SPEDAS workflow skill. .agents/plugins/marketplace.json exposes it as a repo-scoped local marketplace entry.

Both wrappers launch the same spedas-mcp MCP server and share the CDAWeb/PDS/SPICE workflow contract. Validate them with:

python scripts/validate_plugin_packages.py

Claude Code plugin path

A future Claude Code plugin can package this MCP server plus:

  • skills: heliophysics workflow guidance;
  • commands: /spedas:cdaweb, /spedas:pds, /spedas:spice, /spedas:overview;
  • hooks: guardrails around huge downloads, missing output paths, and array-in-chat behavior.

The MCP layer should stay useful without Claude Code: any MCP-compatible harness can connect to spedas-mcp.

Ver o original no GitHub

Bom para

Relacionados