Namashkar Bharat

There are no API keys, accounts or sandbox. Everything here is public and read-only: the directory is the same data every visitor sees, nothing you send is stored, and there is nothing to write to, so there is nothing a key would protect or a sandbox would isolate. Use production directly.

Directory API

GET https://namashkarbharat.com/api/directory

The whole directory in one JSON response: { count, categories[], sites[] }. No parameters and no paging. CORS is open to every origin, and no credentials are used. The OpenAPI 3.1 description has the full schema.

id — Stable slug of the site name, e.g. passport-seva. Changes only if the site is renamed.
name — Site name (English).
url — Official URL on the owning body's own domain.
domain — Hostname without www.
category — One of the 18 category ids (listed in categories[]).
desc — One-line English description.
tags — Lowercase keywords, including common Hindi/Hinglish terms.
org — Owning ministry, department or body (may be empty).
state — ISO 3166-2:IN code without "IN-" (e.g. MH) for sites that serve one state or union territory; absent for national sites.

curl

curl -s https://namashkarbharat.com/api/directory \
  | jq '.sites[] | select(.tags | index("passport")) | {name, url}'

JavaScript (fetch)

const res = await fetch("https://namashkarbharat.com/api/directory");
if (!res.ok) {
  const problem = await res.json(); // application/problem+json
  throw new Error(`${problem.code}: ${problem.detail}`);
}
const { sites } = await res.json();
const official = new Set(sites.map((s) => s.domain));
console.log(official.has("passportindia.gov.in")); // true

Python (standard library)

import json, urllib.request

with urllib.request.urlopen("https://namashkarbharat.com/api/directory", timeout=10) as r:
    directory = json.load(r)

for site in directory["sites"]:
    if site.get("state") == "MH" and "land" in site["tags"]:
        print(site["name"], site["url"])

MCP connector

AI agents can use the remote MCP server at https://namashkarbharat.com/mcp (Streamable HTTP, no sign-in) to find the official site for a need, check whether a URL is official, follow step-by-step playbooks and check portal status. Setup for Claude and other clients, and agents.md, are on the AI agents page.

For agents reading the site

/agents.md: instructions for agents that use this site (workflow, tools, human checkpoints, hard rules).
/llms.txt and /llms-full.txt (every directory entry).
Every page returns Markdown with Accept: text/markdown, or at <page>.md, for example /index.md.
/.well-known/api-catalog (RFC 9727) lists the API with its OpenAPI description and this page.
schema.org JSON-LD in every page, and WebMCP tools in browsers that support WebMCP.

Rate limits

GET /api/directory — No per-client limit: it is a static file served from the CDN. It only changes when the directory is updated, so fetch it once and cache it (refreshing daily is plenty); don't poll.
/mcp (MCP connector) — 120 requests per minute per network (higher for known agent platforms), plus 12 portal_status checks per minute and 60 report_outcome calls per hour. Over the limit: HTTP 429 or a tool error.
/llms.txt, /llms-full.txt, /openapi.json — Static files served from the CDN. No limit.
/api/search, /api/feedback — Not for third-party use (see below): 10 smart searches per network per day, 1 per 3 s and 6 per minute; 60 feedback requests per minute.

Caching

/api/directory — Cache-Control: public, max-age=3600, s-maxage=86400, stale-while-revalidate=604800
/openapi.json, /llms.txt, /llms-full.txt — Cache-Control: public, max-age=3600, s-maxage=86400
/mcp and every error response — Cache-Control: no-store

Errors

Every error from an /api endpoint is RFC 9457 application/problem+json: type (a link to the entry below), title, status, detail, plus code (stable, machine-readable) and hint (what to do next). Unknown /api paths return a JSON 404, unsupported methods a 405 with an Allow header, and any other path returns a JSON 404 when the client prefers application/json. /mcp answers with JSON-RPC errors instead.

HTTP/1.1 404 Not Found
Content-Type: application/problem+json

{
  "type": "https://namashkarbharat.com/developers#errors-not_found",
  "title": "Not found",
  "status": 404,
  "detail": "There is no API endpoint at this path.",
  "code": "not_found",
  "hint": "See /developers for the available endpoints and /openapi.json for their description."
}

`bad_request` · 400 Bad request

The body isn't the JSON object the endpoint expects (malformed JSON, `null`, an array, unknown fields or a filled-in honeypot).

Hint: Send a JSON object with only the documented fields.

Returned by: /api/search, /api/feedback

`too_short` · 400 Question too short

The question is missing or shorter than 3 characters after cleaning.

Hint: Describe what you need in a few words.

Returned by: /api/search

`too_long` · 400 Question too long

The question is longer than 100 characters.

Hint: Keep the question under 100 characters.

Returned by: /api/search

`links` · 400 Links are not accepted

The question contains a URL or `www.`.

Hint: Describe the need in words. To check whether a URL is official, use the MCP connector's verify_official tool.

Returned by: /api/search

`gibberish` · 400 Not a question

The question has too few letters or long runs of repeated characters.

Hint: Use words that describe what you need.

Returned by: /api/search

`forbidden` · 403 Cross-origin request refused

The endpoint only serves namashkarbharat.com's own pages (Origin / Sec-Fetch-Site check).

Hint: Use GET /api/directory (public, CORS-enabled) or the MCP connector at /mcp instead.

Returned by: /api/search, /api/feedback

`verify` · 403 Human check failed

The Cloudflare Turnstile token was missing or invalid. The daily quota unit is refunded.

Hint: Refresh the page and try again.

Returned by: /api/search

`unknown_search` · 403 Unknown search

The searchId/sig pair wasn't issued by /api/search today.

Hint: Send feedback only for a searchId and sig returned by /api/search on the same day (IST).

Returned by: /api/feedback

`not_found` · 404 Not found

There is no endpoint or page at this path.

Hint: See /developers for the available endpoints and /openapi.json for their description.

Returned by: any /api path; any path when the client prefers application/json

`method_not_allowed` · 405 Method not allowed

The endpoint exists but doesn't support this HTTP method.

Hint: Use one of the methods listed in the Allow header.

Returned by: every /api endpoint

`too_large` · 413 Request too large

The body is over the endpoint's limit (2 KB for /api/search, 1 KB for /api/feedback).

Hint: Send a smaller body; only the documented fields are needed.

Returned by: /api/search, /api/feedback

`unsupported_type` · 415 Unsupported media type

The body's Content-Type isn't application/json or text/plain.

Hint: Send Content-Type: application/json.

Returned by: /api/feedback

`burst` · 429 Too many searches too quickly

More than 1 search in 3 seconds or 6 in a minute from one network, or a search sent less than 0.8 s after the page loaded.

Hint: Wait a few seconds (see Retry-After) and try again.

Returned by: /api/search

`daily` · 429 Daily smart-search limit reached

This network has used its smart searches for today. Cached answers still work.

Hint: The limit resets at midnight IST (see resetIn and Retry-After, in seconds). Browse /directory or use /api/directory meanwhile.

Returned by: /api/search

`rate_limited` · 429 Too many requests

More than 60 feedback requests in a minute from one network.

Hint: Slow down and retry after a minute.

Returned by: /api/feedback

`server_error` · 500 Server error

Something unexpected failed on our side. Nothing was stored.

Hint: Retry later. If it keeps happening, report it on X (@ajinkya_shejul).

Returned by: every /api endpoint

Not for third-party use

/api/search (the site’s AI-assisted search) and /api/feedback only accept requests from namashkarbharat.com’s own pages; other origins get 403 forbidden. They are left out of the OpenAPI description on purpose. Use the directory API or the MCP connector instead.

Fair use

Free to use, with no key or sign-up. Please cache /api/directory instead of fetching it on every request.
Send people to the official website itself, never to a copy, a mirror or a paid intermediary, and don't mix unofficial sites into anything you present as this directory.
Don't imply that Namashkar Bharat, or your product, is a government service or endorsed by the government. We are an independent guide.
Leave logins, OTPs, payments and final submissions on official portals to the person. Agents must hand over at those steps.
No uptime guarantee. The directory is maintained by one person and checked regularly, but government sites move; report problems on X.

Privacy

No keys, accounts or cookies on the API, so there is nothing that identifies you.
Never send personal data (Aadhaar, PAN, phone numbers, passwords, OTPs) to any endpoint here. None of them need it.
GET /api/directory and the static files are plain CDN responses. We add no logging of our own; the hosting platform keeps its standard short-lived request logs.
The MCP connector doesn't log search words, checked URLs or tool arguments; rate limits use a salted hash of the network address, never the address itself.

API changelog

Unreleased: The agent setup page moved from /connect to /agents (permanent redirect), and /agents.md serves instructions for agents. Errors from every /api endpoint are RFC 9457 application/problem+json with a stable code and hint; unknown /api paths return a JSON 404 and unsupported methods a 405 with Allow. This developer page, /.well-known/api-catalog (RFC 9727) and service-desc/service-doc links. Directory sites that serve one state carry a state field.
2026-09-30: GET /api/directory opened to other origins (CORS, categories[] added), /openapi.json, /llms.txt, /llms-full.txt, Markdown versions of every page, and the MCP connector at /mcp.

Contact

Questions, broken links and data requests: @ajinkya_shejul on X. Namashkar Bharat is an independent guide, not a government website (about).

Build on the official directory

Directory API

curl

JavaScript (fetch)

Python (standard library)

MCP connector

For agents reading the site

Rate limits

Caching

Errors

bad_request · 400 Bad request

too_short · 400 Question too short

too_long · 400 Question too long

links · 400 Links are not accepted

gibberish · 400 Not a question

forbidden · 403 Cross-origin request refused

verify · 403 Human check failed

unknown_search · 403 Unknown search

not_found · 404 Not found

method_not_allowed · 405 Method not allowed

too_large · 413 Request too large

unsupported_type · 415 Unsupported media type

burst · 429 Too many searches too quickly

daily · 429 Daily smart-search limit reached

rate_limited · 429 Too many requests

server_error · 500 Server error

Not for third-party use

Fair use

Privacy

API changelog

Contact

`bad_request` · 400 Bad request

`too_short` · 400 Question too short

`too_long` · 400 Question too long

`links` · 400 Links are not accepted

`gibberish` · 400 Not a question

`forbidden` · 403 Cross-origin request refused

`verify` · 403 Human check failed

`unknown_search` · 403 Unknown search

`not_found` · 404 Not found

`method_not_allowed` · 405 Method not allowed

`too_large` · 413 Request too large

`unsupported_type` · 415 Unsupported media type

`burst` · 429 Too many searches too quickly

`daily` · 429 Daily smart-search limit reached

`rate_limited` · 429 Too many requests

`server_error` · 500 Server error