Connect your agent to MCPOrbital.
One Apache-2.0 MCP server. 8 grounded tools across NASA JPL SBDB / CAD / Horizons, CelesTrak, NOAA SWPC + DSCOVR, NASA Exoplanet Archive, and arXiv. Every response carries a provenance[] chain to the primary source. Works with Claude Desktop, Cursor, Cline, the Anthropic / OpenAI / Gemini SDKs, or plain HTTP.
- 1
Install
npx -y @mcporbital/space-intelligence-mcp@latestNo global install needed. The MCP server runs on-demand; most clients spawn it as a subprocess.
- 2
Configure your client
Pick your client below — Claude Desktop, Cursor, Cline, or any agent SDK. Paste the JSON / Python.
- 3
Verify + first query
Ask “research brief on Apophis” — MCPOrbital chains the right tools and cites every value.
Most common path. Edit the JSON config file once; restart Claude Desktop; the skill activates on any space-research trigger.
- macOS · ~/Library/Application Support/Claude/claude_desktop_config.json
- Windows · %APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"space-intelligence": {
"command": "npx",
"args": ["-y", "@mcporbital/space-intelligence-mcp@latest"],
"env": {
"NASA_API_KEY": "DEMO_KEY"
}
}
}
}After savingQuit and reopen Claude Desktop (Cmd-Q, not just close). The Settings → Developer panel should show `space-intelligence` with 8 tools advertised.
“Give me a research brief on 99942 Apophis — orbit, physical properties, close approaches, latest papers.”
MCPOrbital will chain search_space_object → get_object_profile → get_asteroid_close_approaches → get_ephemeris, citing every value.
What your agent can ask
Same tool surface in Claude Desktop, the SDK, and the REST API. Every response includes a provenance[] envelope.
| Tool | Sources | What it returns | Example input |
|---|---|---|---|
| search_space_object | SBDB · MPC | Resolve free-form → canonical id | { "query": "Apophis" } |
| get_object_profile | SBDB · arXiv | Orbit + physical + discovery + papers | { "primary_id": "99942 Apophis" } |
| get_ephemeris | JPL Horizons | OBSERVER (RA/Dec/Mag) or VECTORS (state) | { "primary_id": "99942", "center": "I11" } |
| get_asteroid_close_approaches | JPL CAD | Past + future close approaches | { "primary_id": "99942", "dist_max_au": 0.05 } |
| get_space_weather_now | NOAA SWPC · DSCOVR | Kp / Bz / solar wind / X-ray snapshot | {} |
| get_satellite_tle | CelesTrak GP | Latest TLE for any catalogued sat / debris | { "norad_id": 25544 } |
| get_exoplanet_data | NASA Exoplanet Archive | ADQL query against ps / pscomppars / toi / k2pandc | { "hostname": "TRAPPIST-1" } |
| compare_data_sources | SBDB (v0.2: + MPC + ADS) | Cross-source diff on a named scalar | { "primary_id": "99942", "field": "physical.diameter" } |
Common questions
I get UNABLE_TO_GET_ISSUER_CERT_LOCALLY for JPL endpoints
brew install ca-certificates export NODE_EXTRA_CA_CERTS="$(brew --prefix)/etc/ca-certificates/cert.pem"Or run the MCP server in Docker — the Linux base image has fresh roots.
How do I get a personal NASA_API_KEY?
ADS / Space-Track integration?
How do I verify the server is healthy?
~/.config/claude/skills/mcporbital-space-research/scripts/verify_mcp.shExpected: ✓ @mcporbital/space-intelligence-mcp reachable. 8 tools advertised.
Self-host the MCP server?
Streamable HTTP / hosted endpoint for claude.ai with managed MCP?
Ready to ground your agent?
Open the Apophis demo to see what your scientist will see, or jump into the GitHub source to read the safety layer line-by-line.