An MCP server that lets an AI agent search ISRO's Bhoonidhi archive in plain language, download the scenes that are ready, and stage the rest to your cart.
uvx --from git+https://github.com/geovicco-dev/bhoonidhi-mcp bhoonidhi-mcpSearching the old way means knowing the satellite token, the bounding box, and the date format up front. bhoonidhi-mcp lets the agent fill those in from what you said.
The agent resolves “Shillong” to a bounding box, matches “ResourceSat-2A” to the portal’s platform tokens, and runs the live search — then reports which results are actually downloadable.
One conversation covers it end to end: search and find open data, download the scenes that are ready, stage priced or on-order scenes to your cart, and check a saved search for anything new. Pick a flow — the agent calls the tools, and the JSON they return is what this page turns into the map and summary on the right.
The server returns structured JSON — scene coordinates and availability — not a picture. The map and summary here are this page's illustration of that data. Satellite names, tokens, and coordinates are real; scene IDs and counts are representative of a live query.
Each is a thin wrapper over the same client the bhd command line uses. Search, preview, and saving a search are open to anyone. Downloading and cart use your own Bhoonidhi login — never this project's.
bhoonidhi-mcp runs over stdio: the client launches it as a subprocess and talks over stdin and stdout. Every setup comes down to “run this command.”
You log in to Bhoonidhi yourself, outside the agent — with bhd auth login in your terminal, or credentials in the server's environment. Your password never reaches the agent. All it can check is whether a valid session exists, through auth_status.
bhoonidhi-mcp does not talk to the portal directly. It calls the same client the bhd command line uses, so there is one place where portal logic lives.
The SDK and bhd CLI: portal grammar, satellite tokens, availability classification, and downloads. bhoonidhi-mcp depends on it.
Turns natural language into structured portal queries and shapes the results for an agent. It adds the geocoder and the fuzzy satellite matching — nothing about the portal itself.
Availability is reported honestly. The server classifies every scene the same way the CLI does — Ready, Archived, On order, or Priced — so an agent can tell you what you can actually download before you try, instead of guessing from the scene name.
Why it exists. The Indian Space Policy 2023 opened much of ISRO's Earth-observation archive for everyone to use. Having built the bhd CLI and SDK to reach that open data from code, I wanted to take it a step further — so an agent could reach it from plain conversation too. The sole motive is to make this data easier to access and to put it within reach of anyone working with Indian Earth observation data; the roadmap follows what people actually need.
The imagery reached through this server belongs to ISRO/NRSC and is governed by the Bhoonidhi EULA, not by this project's MIT license — that license covers the code, nothing else. You download with your own Bhoonidhi account; open data is free to use, publish, and build on, but the original products can't be redistributed commercially in their original form, and priced or on-order scenes are completed on the portal. Anything you publish from the data must credit it as ISRO-IRS.
uvx --from git+https://github.com/geovicco-dev/bhoonidhi-mcp bhoonidhi-mcpNeeds uv. The first run fetches and builds from source; later runs start from cache.