Best for
- Farmers, city planners, NGOs, and students without GIS expertise
- Developers who want natural-language satellite data analysis
- Teams wanting zero-credential self-hosting
A
Axion Planetary MCP
Community
The world's first virtual satellite you can connect with MCP — Earth observation as easy as conversation.
- Category
- Other #150 of 230
- Stars
- ★ 221 Popular
- Transport
- SSE
- Runtime
- Node.js 18+ · Docker · Hosted (no local runtime)
- Credentials
- Optional API key
- License
- MIT
- Last commit
- ⚠
- Tools
- 6
50FMRS · D
Positioned as a 'virtual satellite,' this project lowers the barrier to Earth observation with end-to-end tools for search, mapping, processing, classification, and export over free public data, requiring no cloud credentials when self-hosted — great for rapid prototyping and non-experts; however, the hosted free tier is limited, and advanced features like SAR-to-optical are hosted-only.
Strongest · Documentation 15/20 Weakest · Security and permissions 3/20
Why each score
- Reliability 9/20
- The README documents 6-7 tools (axion_data, axion_map, axion_process, axion_classification, axion_export, axion_system, plus the hosted-only axion_sar2optical) and mentions a /health endpoint, showing some basic observability intent. However, no manifest was cached and the README contains no automated tests, CI workflow, or reproducible integration-test evidence to confirm that declared tool behavior matches the actual implementation, nor is error handling documented beyond a short troubleshooting list. Per the static calibration rule, reliability cannot exceed 12 without CI+test evidence, so it is scored 9: the happy path (npx install, SSE bridge) appears plausible, but edge cases, auth failures, and upstream STAC API timeouts have no documented, diagnosable handling.
- Security and permissions 3/20
- Red line triggered: the README publishes a working 'starter' API key (axion_6C1D_634DDAEAA324235F) directly in the documentation that anyone can copy and use against the hosted service — this matches the red-line criterion of install examples containing real tokens/secrets. Additionally, the hosted path routes all of a user's location/imagery queries through a third-party server run by a single individual (axion-mcp-sse.onrender.com) with no disclosed data retention, logging, or privacy policy. While the env-var reference table claims REQUIRE_AUTH defaults to true, the very first Docker quick-start example in the same README runs with `REQUIRE_AUTH=false`, i.e., the demonstrated default configuration is an unauthenticated server exposed on port 3000 with no warning or confirmation step. These combined factors constitute an unresolved red-line risk, so security is capped in the 0-4 band and scored 3.
- Maintenance 10/20
- The repository declares an MIT license and is not archived, and has 222 stars with 0 open issues — the latter could reflect either good triage or low engagement, and is not treated as a positive maintenance signal on its own. The README contains no release history, changelog, roadmap, or security-response channel, and no dependency-update policy is described; with the manifest uncached, actual version numbers, recent commit activity, or dependency freshness cannot be verified. Overall this reflects basic governance elements (license, non-archived status) without sufficient evidence of active, ongoing maintenance, scored 10.
- Documentation 15/20
- Documentation is fairly thorough: step-by-step Claude Desktop configuration for multiple OSes, an environment-variable reference table, Docker/Docker Compose examples, an architecture diagram, a data-source resolution/coverage table, and a troubleshooting section — these substantially lower the onboarding barrier. Deductions: no per-tool input parameter schema/JSON spec is provided, so parameters cannot be verified against the actual implementation; cost/rate-limit/free-tier constraints are only briefly mentioned ('sleeps for 30 seconds'); and no data-privacy or retention policy is disclosed. Scored 15.
- Setup experience 13/20
- The hosted version offers a clear two-minute setup flow with multiple installation options (npx, global install, direct sse-bridge.mjs invocation) and cross-platform config examples; the self-hosting section also provides detailed Docker and from-source commands. However, per the static calibration rule, setup cannot exceed 15 without CI and test evidence. The hosted service also runs on Render's free tier with a known cold-start delay ('may take 30 seconds to wake up'), and obtaining a personal API key requires emailing the maintainer or filing a GitHub issue (the shared demo key may be rate-limited or contended), introducing some fragility. Scored 13.
Static review · not runListed 2026-09-07
Read the FMRS scoring method →Fit and risk
What it can accessUses the network
Not for
- Users needing high-resolution commercial imagery
- Users unwilling to self-host but wary of the free hosted tier
- Production remote-sensing workflows requiring sub-meter real-time data
Required permissions
- API key (for hosted server authentication)
- Optional AWS credentials (only for GeoTIFF exports to S3)
- Local network access to reach the SSE server
Risks and side effects
- The hosted server runs on Render's free tier and may take ~30 seconds to wake when idle
- The starter API key is published in the README and may be abused or revoked
- Exports over 10MB require an S3 setup
- AWS credentials for S3 exports must be kept secure
Setup
Before you start
Runtime：Node.js 18+ · Docker · Hosted (no local runtime)
AXION_API_KEY
API key for the hosted Axion server; a starter key is in the README, or request one via [email protected] or a GitHub issue
AWS_ACCESS_KEY_ID
AWS key needed only for exporting GeoTIFFs to S3
AWS_SECRET_ACCESS_KEY
AWS secret needed only for exporting GeoTIFFs to S3
Other optional settings (4)
REQUIRE_AUTH
Whether self-hosted server requires API key auth; defaults to true
PORT
Self-hosted server port, default 3000
AWS_REGION
AWS region, defaults to us-east-1
AXION_S3_EXPORTS_BUCKET
S3 bucket name for storing GeoTIFF exports
- Get an API key (try the starter key in the README or request one by email).
- Add an mcpServers entry to your Claude Desktop config using npx -y axion-mcp --api-key YOUR_API_KEY.
- Restart Claude Desktop. Alternatives: global npm install of axion-mcp, or self-host via Docker (port 3000, optional REQUIRE_AUTH) and connect with sse-bridge.mjs to http://localhost:3000/sse.
claude_desktop_config.json
{"mcpServers":{"axion":{"command":"npx","args":["-y","axion-mcp","--api-key","YOUR_API_KEY"]}}}
Shown for Claude Desktop. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.
.vscode/mcp.json
{
"servers": {
"axion": {
"command": "npx",
"args": [
"-y",
"axion-mcp",
"--api-key",
"YOUR_API_KEY"
]
}
}
}
Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).
Terminal
claude mcp add axion -- npx -y axion-mcp --api-key YOUR_API_KEY
Run it in a terminal; replace any <…> placeholders with your own values first.
Check that it works
Restart Claude Desktop and check that axion_data, axion_map, axion_process, axion_classification, axion_export, and axion_system appear in the tool list; asking for an NDVI vegetation map should return a result to confirm the connection.
Troubleshooting
- 'npx not found': install Node.js 18+
- 'Unauthorized': verify the API key has no extra spaces or use the AXION_API_KEY env var
- 'Connection refused': the hosted server may be sleeping; wait 30 seconds or self-host
- Tools not showing in Claude: fully quit and restart Claude Desktop, and validate the config JSON syntax
Things to try
Once connected, you can ask your AI assistant things like:
- Show me NDVI vegetation health for Iowa farmland
- Calculate the water index for Lake Mead
- Create a false color composite of the Amazon
- Classify land cover in California's Central Valley
Tools 6
axion_data read-only
Search satellite imagery from Sentinel-2, Landsat 8/9, and NAIP collections.
axion_map read-only
Create interactive maps with NDVI/EVI/SAVI, NDWI/MNDWI, NDBI indices and true/false color composites.
axion_process read-only
Process imagery: cloud-free composites, temporal mosaics, 9+ spectral indices, band math.
axion_classification read-only
ML land cover classification using Random Forest with custom training points and accuracy metrics.
axion_export writes
Export GeoTIFF, PNG thumbnails, XYZ map tiles, and S3 cloud storage (S3 requires AWS credentials).
axion_system read-only
System utilities: health check, configuration info, available collections.
No matching tools
Use cases
Monitor crop health on farmland with NDVI
Track water quality and levels in lakes and reservoirs
Analyze urban expansion and heat islands
Monitor forest health and real-time deforestation
Supported clients
Claude Desktop
Listed from the project's documentation, not tested by this site.
Overview
Axion Planetary MCP (MIT-licensed) is a foundation for democratizing geospatial AI agents, fully rebuilt on AWS in V2.0. It lets users search and analyze free public satellite imagery (Sentinel-2, Landsat, NAIP) via natural language over MCP, with data sourced from the AWS Earth Search STAC API (no auth required). It offers a hosted SSE server (via an npx bridge) and self-hosting via Docker or source; basic self-hosted functionality needs no cloud credentials.
Similar servers
Hevy MCP
77 · B
Manage your Hevy workout data from AI assistants via MCP.
Firecrawl MCP Server
75 · B
Firecrawl's official MCP server for web search, scraping, and structured extraction for AI agents
BioMCP
75 · B
One binary. One grammar. Evidence from the biomedical sources you already trust.
Compartment
75 · B
Encrypted, fully offline long-term memory for AI agents
Source revision 97667b59e53d Data synced 2026-10-19 Read the FMRS scoring method