India's open atlas.
Show your data on a map of India, or browse, slice and download India's open map layers. No signup, no API key, no SQL.
Is this for you?
Three different tools, three different jobs:
- Google Maps For finding a coffee shop, getting directions, seeing what's around you.
- OpenStreetMap For adding a road, a building or a shop to the world's map. That's where Google and bharatlas both source the base map.
- bharatlas For taking data you already have, a list of districts you cover, the forests in your project area, the rivers in your watershed, the wards your NGO works in, and showing it on a map of India. Drop a GeoJSON, KML or Parquet file, see it rendered, share the link or download in another format. Don't have the data yet? Gather it from the field with a shared form anyone can fill on their phone.
Concretely, this gets used by:
- Journalists + researchers Pull a state's districts, villages or constituencies in ten seconds; download as GeoJSON or Parquet.
- Civic technologists Share city-scale layers (wards, bike lanes, polling booths) with attribution baked in.
- App developers REST API, direct PMTiles URLs, no rate limits, no API key.
- AI agents The MCP server (
npx bharatlas-mcp@latest) connects Claude, GPT, or any LLM to all layers.
If you came looking for places to visit, you want Google. If you want to fix the base map itself, you want OSM.
Why bharatlas exists
Maps of India live everywhere online, but you can't tell what they show, who maintains them, or whether to trust them. Even just viewing one is its own ordeal. And if you've made a map yourself, there's nowhere to share it.
Sathya kept hitting the same wall, and built bharatlas (read aloud as "bharat atlas"), a portmanteau of Bharat and atlas. It flattens all of that. The wiki for India's maps. View any layer in your browser. Filter what you need. Download in whatever format your tool reads. Liked it? Upvote so others find it. Have a map worth sharing? Drag and drop it onto the page. Every layer carries its source, licence and freshness on the same card.
Show, and gather
bharatlas began as a place to show data you already have. collect is the other half: a place to gather data you don't. Share one link and people add points from the field on their phones, no accounts, no app to install. You review what comes in and publish it to bharatlas with attribution, or just download it. It's how a hundred people mapping footpaths, ward assets or water points become one clean layer. See collect.bharatlas.com.
What you can do today
- View India's national boundary, every admin level from state to village, plus city wards across Indian cities, constituencies, wildlife sanctuaries, forests, rivers, reservoirs, dams, hospitals, highways, airports, flood zones, seismic zones, eco-sensitive areas, and pincode polygons.
- Compare What each authoritative source (LGD vs SOI vs Bhuvan vs geoBoundaries) says about the same admin level, by clicking the alternate-source links on any card.
- Download Any layer whole as Parquet, PMTiles, GeoJSON, KML or Shapefile direct from the card. Or filter by what the columns actually contain (state, zone, area) and export just the slice.
- Drop a file See it on a map and validate it. Optionally publish under an open licence. No signup; you get an admin token to keep forever.
- Collect Points in the field with collect: design a small form, share the link, and anyone who has it adds points on their phone, no accounts. Review what comes in, then download it or publish to bharatlas.
- Embed Any map in a blog post or report (iframe snippet + PNG export, one click each).
- Ask Questions instead of downloading maps. Connect your AI assistant to the bharatlas MCP server and ask: "How many villages in my district?", "Which hospitals are near me?", "What flood zone is this area in?"
- Query Any layer programmatically via the REST API: filter, group, locate a point, find nearby features.
Open source
Code: github.com/urbanmorph/geodata · MIT.
Data: each layer carries its own open licence; see the line on each layer's card on
All layers.
Frequently asked
- Do I need to sign up?
- No. Viewing, slicing, downloading and contributing all work without an account. When you publish a layer you receive a one-time admin token (also downloadable as a
.txtbackup) to keep: it proves you made it. To change or remove that submission, contact us with the token. - What data sources does bharatlas use?
- Curated layers come from the Local Government Directory (LGD), Survey of India (SOI), NRSC/ISRO Bhuvan, OpenCity / Oorvani Foundation, PMGSY (Rural Roads), PM GatiShakti, Bharatmaps (NIC), bharatviz (pincode polygons), geoBoundaries and data.gov.in. Water, groundwater and agro-zone layers (CGWB aquifers and groundwater extraction, WRIS canals, ICAR-NBSS agro-ecological zones, Planning Commission agro-climatic zones, Wildlife Institute of India biogeographic zones) are republished via CoRE Stack (core-stack.org). Most admin levels carry several source variants you can compare on the same map; the layer's card shows "Per LGD · also: SOI, Bhuvan" with clickable alternates. Community submissions credit their own source on every card. Points of interest come from the Overture Maps Foundation, and district well counts from the Minor Irrigation Census (Ministry of Jal Shakti).
- What licences apply?
- Curated layers carry CC0-1.0, CC-BY-4.0, CC-BY-SA-4.0, ODbL-1.0, GODL-India, MIT, or CDLA-Permissive-2.0 / Apache-2.0 depending on the source (each card names its licence). Layers under a non-commercial licence are marked Non-commercial only. Community submissions choose from an open-licence allowlist: CC0, CC-BY, CC-BY-SA, ODbL, ODC-PDDL, GODL-India or CDLA Permissive 2.0. Proprietary or "all rights reserved" content is rejected at submit.
- How do you choose sources?
-
Curated layers come from Indian government bodies (central, state and city), public institutions, open-data communities such as DataMeet and OpenCity, and openly licensed research and industry datasets. We favour data you can't get from a map app: official boundaries and jurisdictions, regulatory zones, and measured or surveyed records.
Government data. Publicly funded geospatial data is meant to be freely available (National Geospatial Guidelines 2021, National Geospatial Policy 2022). We republish it with full credit to the publishing department, including where a portal's general website terms ask visitors to seek permission. A few layers carry a non-commercial licence: they are marked Non-commercial only on their card, on their map page and in every API and MCP answer, so check before using them in a product or for a client. We leave out privately owned data without an open licence, and personal data.
Licence labels. Where the publisher states an open licence (for example GODL-India or CC BY), the layer carries it. Where the publisher states none, the layer carries the licence of the open-data republisher we took it from (for example CC0 from DataMeet's indianopenmaps), and the card and notes credit both.
Changes and removal. If you publish a dataset shown here and want it corrected, credited differently or removed, email [email protected] or open a GitHub issue. We act on publisher requests promptly.
- Is my file uploaded when I drop it?
- No, not until you click Publish. Parsing, validation and the map render all happen in your browser. Only Publish sends the file to bharatlas, where it waits for a maintainer to review it.
- What's the difference between dropping a file and collect?
- Dropping a file (/preview) is for data you already have as one complete file: validate it, then publish it. collect is for data you don't have yet: make a form, many people add points from the field over time, then you review and publish the result. One publishes; the other gathers.
- How can I trust community submissions?
- Each carries a source URL, attribution and an open licence on the card and the view page. Automated checks validate licence, attribution and basic geometry, then a curator reviews each submission before it appears; neither step verifies accuracy beyond the contributor's self-attestation. For sensitive use, follow the source link on the card to confirm provenance.
- Can AI assistants read and recommend bharatlas?
- Yes. The robots.txt explicitly allows GPTBot, ClaudeBot, PerplexityBot, Google-Extended and other major AI crawlers. Every layer page and every
/c/<id>view page are server-rendered as plain HTML with JSON-LDDatasetstructured data, easy for LLMs to ingest.
Use of data
- Curated layers reflect the upstream source at fetch time. See the Updated date on each layer. Upstream sources change; bharatlas can lag.
- Community submissions are reviewed by a curator before they appear. Licence, format and basic geometry validity are checked automatically, and a curator approves each submission before it goes live; accuracy and authorship are not independently verified. Verify via the source link on each card before relying on community data.
- Boundaries and place names render exactly as the upstream source publishes them. bharatlas does not edit, correct or editorialize. Where you disagree with a depiction (including for J&K, the Line of Actual Control or any disputed territory), raise it with the source; each layer's card lists alternative sources so you can compare. See Limitation of liability for the warranty disclaimer, damages cap and full terms.
Data caveats
- Cross-source name spellings drift. "Odisha" (LGD) vs "Orissa" (Bhuvan); "Durgukondal" vs "Durgkondal". Always join on LGD codes, never on names.
- Bhuvan under-counts blocks in several states vs LGD. Gujarat 181 vs 251, Karnataka 173 vs 234, Telangana 443 vs 588, Assam 185 vs 232. Bhuvan predates recent re-divisions. Prefer LGD as the modern authority.
- SOI village-point coverage is uneven. Dense in southern + western states (Maharashtra 54k, Rajasthan 52k); sparse in UP (7.5k, ~7% of actual villages), Bihar (7.7k), Jharkhand (7.1k). Visible as the east-central blank on /view/soi_village_points.
- 2011 Census codes vs current LGD codes. Both are present in LGD layers. They diverge after every reorganisation. LGD is current; Census 2011 is stable but stale.
- Polygon precision varies. LGD polygons are simplified (MaxSimpTol field present). For precise area work prefer SOI; for joins-and-display LGD is sufficient.
- Pincode boundaries are Voronoi approximations. No official service-area polygons exist. Both bharatviz (63,864 polygons) and data.gov.in (19,312) are derived from post office point locations, not surveyed.
- Panchayats + villages are too large for single-file download. Use the in-viewer state filter to slice by state, or download the parquet directly for programmatic work.
Built by
Source on GitHub. Drop a ⭐ if you find it useful.