Pricing

Successful calls to a metered tool — the property, locality and decision tools. Failed calls are not counted. A cached response does count: you asked for the data and got the data, and the cache is our optimisation rather than your discount. Repeating a query you have already made is the cheapest thing you can do on this API, but it is not free.

It depends on the tier.

• Developer: 1,000 calls / month is the limit. Calls above it are refused with HTTP 429 (on REST, error: "rate_limited", reason: "usage_month") and a message naming the reset date and the tier that carries more. Nothing is deleted. Before it, a call can also be refused by the per-minute rate or the fair-use daily line.
• Business: 10,000 calls / month included. Calls above the allowance are served, not refused, and billed at the tier’s own per-call rate — then ₹0.45 per call + GST, up to 30,000 calls / month.
• Scale: 50,000 calls / month included. Calls above the allowance are served, not refused, and billed at the tier’s own per-call rate — then ₹0.30 per call + GST, up to 1,50,000 calls / month.
• Enterprise: volume, and anything above it, is set by your contract.

On Business and Scale, calls stop only at the ceiling, 3× the monthly allowance: past it a REST call returns HTTP 429 with error: "rate_limited" and reason: "usage_ceiling", and an MCP call a tool error naming the ceiling and the reset date, until the month resets at 00:00 UTC on the 1st. The ceiling is a contract term and can be raised for an application on request. Overage is totalled per UTC calendar month and invoiced with GST, on usage from October 2026 (UTC) onward; an earlier month stays at the flat monthly price, with no ceiling. Every REST response carries X-AVnester-Usage-Limit, X-AVnester-Usage-Remaining and X-AVnester-Usage-Reset for your application’s monthly allowance, so you can build against it rather than discover it (and, on a tier that bills overage, X-AVnester-Usage-Overage and X-AVnester-Usage-Ceiling), plus the IETF RateLimit-Policy and RateLimit headers for the per-minute budget (RateLimit-Policy: "<tool>";q=<limit>;w=60, RateLimit: "<tool>";r=<remaining>;t=<seconds>). A REST call made without signing in also lists its per-IP budgets across all tools, per minute and per UTC day. Every 429 carries error: "rate_limited", a reason naming the limit (rate_limit_per_tool, anon_aggregate, anon_daily, cache_miss_budget, locality_cardinality, usage_daily, usage_month, usage_ceiling) and a Retry-After header: seconds to the end of the one-minute window the first call opened for a per-minute limit, to 00:00 UTC for a daily one, and to the month’s reset for the monthly allowance or the ceiling. Over MCP the refusal is a tool error whose message names the limit, except the per-IP limit across all tools for calls made without signing in, which answers before any tool runs with the same HTTP 429 as REST. On MCP the transport commits the response headers before the tool runs, so an authenticated tools/call carries the X-AVnester-Usage-* allowance headers as they stood when the request started and no per-tool RateLimit-Policy/RateLimit pair.

A tenth of your application’s monthly allowance, per UTC day; it resets at 00:00 UTC. It exists because a per-minute rate limit cannot stop a runaway loop from spending a whole month in minutes — at 600 requests a minute (per tool), a 10,000-call month is gone in under twenty. On Developer, calls above it are refused with HTTP 429 (reason: "usage_daily" on REST). Across a 30-day month it adds up to three times the monthly total, so it can never cost you calls within your allowance — it only stops them being spent all at once. On Business and Scale it is an alert, never a refusal: a paying application’s busiest day is served, and the monthly ceiling is what stops a runaway loop.

A cap on how many different localities you query in a UTC day, counted across every tool: 50 a day on Developer. Asking about the same localities repeatedly never trips it — a site covering a fixed set of neighbourhoods will not notice it exists. A call that names a new locality past the cap is refused with HTTP 429 (reason: "locality_cardinality" on REST) until 00:00 UTC; the localities you have already queried that day stay available. Business, Scale and Enterprise have no such cap: walking the whole corpus is what a licence is for.

Developer: 1 active app per account. Its allowance is counted per app, so the limit is what stops a second app from being a second allowance. Revoking an app frees its place; apps you already have are never revoked for you. Business, Scale and Enterprise are set per app: we move the app you name, and the account can then register another Developer app beside it.

Developer is an evaluation tier: build, test, run a personal project, with visible attribution back to AVnester. Redistributing the data, building a derivative dataset, or putting it in front of your own paying customers needs a Business licence. The full terms are in the AI / MCP Acceptable Use policy.

By invoice. Email partnerships@avnester.com and we will raise a GST invoice against your GSTIN and enable the tier on your existing OAuth client — no migration, no new credentials. Calls above the monthly allowance are totalled at the end of each UTC month and invoiced the same way. There is no card checkout on this surface today, deliberately.

No. Both transports run through the same limiter and read the same per-tier number, so a tier means one thing regardless of how you connect. On the OAuth tiers the per-minute budget is per tool: per authorising user for OAuth calls, and per application for server API keys. What differs is the headers: an MCP response cannot carry the per-tool per-minute headers, because the streaming transport sends its headers before the tool runs — read them from the REST endpoint of the same tool, which carries every header on every response.