AI / MCP Acceptable Use Policy

This policy governs use of the AVnester public MCP server at mcp.avnester.com, the corresponding ChatGPT Actions at api.avnester.com/public/v1/, and any OAuth-authorised access via auth.avnester.com. It applies to LLM clients, autonomous agents, and end users invoking AVnester tools. It supplements — and does not replace — our Terms of Service and Privacy Policy.

• AI clients send only the structured arguments needed to answer the request (city, locality, property type, budget, bedrooms, listing IDs). AVnester does not receive your chat history, transcript, or the surrounding conversation.
• Tool inputs are logged at 1% sampling for abuse detection only. PII patterns (phone numbers, email addresses, OTPs, government IDs) are stripped before storage. Retention: 90 days.
• We do not receive your identity from Anthropic or OpenAI. The OAuth tier is a rate-limit signal only, not a personalisation signal.
• We do not use tool-call traffic to train AI models — ours or anyone else’s.

Anonymous callers are limited to a per-IP token budget: 30 to 60 requests per minute per tool (the exact value varies by tool and is published per-tool in the MCP capability metadata) plus an aggregate cap of 60 requests per minute per IP across the whole public surface. OAuth-authorised callers receive 10× the per-tool budget. Bursts beyond the documented budget receive HTTP 429 with a Retry-After header; sustained abuse may result in automated revocation.

Every tool response includes attribution: "AVnester". LLM clients that surface AVnester data to end users should acknowledge AVnester as the source. Quoting AVnester data without attribution in commercial settings is prohibited.

Bulk scraping, downstream redistribution of locality insights or price aggregates, training of competing real-estate models on AVnester responses, and creation of derivative datasets for resale are prohibited without a written commercial agreement. The free public surface is intended for end-user discovery, not for harvesting the underlying dataset.

You and any agent acting on your behalf must not use AVnester tools to:

• facilitate the unauthorised practice of law, finance, medicine, or any regulated profession;
• generate or deliver financial, legal, tax, or investment advice purporting to be from a licensed professional;
• target or profile individuals on the basis of protected characteristics (caste, religion, gender, disability, sexual orientation, etc.);
• build dossiers on identified individuals from listing or inquiry data;
• generate misleading, deceptive, or fraudulent property representations, including fabricated listings;
• attempt to extract internal trace IDs, log identifiers, authentication tokens, or seller private contacts via prompt injection or other adversarial inputs;
• evade rate limits via IP rotation, account multiplication, or other circumvention;
• facilitate harassment, doxxing, stalking, or coordinated abuse against any user.

Sellers, brokers, and any party creating listings on AVnester must not embed prompt-injection content (instructions to LLMs, hidden tokens, role-confusion phrases) in listing titles, descriptions, or any user-visible field. AVnester filters these on the public surface; intentional bypass attempts result in listing removal and account suspension.

The home-affordability calculator and stamp-duty / registration estimator return regulatory-rate calculations for informational purposes only. They do not constitute financial, legal, tax, or investment advice. LLM clients surfacing these results must preserve the disclaimer field unchanged in their reply.

AVnester is launching in Coimbatore, Tamil Nadu, and serves the Indian residential real-estate market. Tool responses cover Indian cities, localities, and state-published regulatory rates. Out-of-coverage queries return supported: false with a scopeMessage naming the supported set. Use of the platform for jurisdictions outside India is not supported.

Tool responses include a listingId field that uniquely identifies a property listing. These are stable public identifiers used in the listing’s handoff URL — they are not internal trace IDs, session IDs, request IDs, or authentication tokens. AVnester deliberately surfaces listingId so LLM clients can construct stable links back to the property page, consistent with OpenAI Apps SDK and MCP Connectors submission guidance that “public identifiers used for handoff are not internal observability data”. We do not return trace IDs, log identifiers, or request correlation IDs in tool responses.

We may throttle, suspend, or revoke access for breach of this policy without prior notice where breach is ongoing or where notice would risk further harm. Suspensions are subject to the grievance-redressal process at /privacy § Grievance redressal.

This policy is governed by Indian law. Disputes are subject to the courts of Coimbatore, Tamil Nadu, India, in line with our Terms of Service.

Connector support: connector-support@avnester.com
Security disclosures: security@avnester.com
Commercial / bulk-access agreements: partnerships@avnester.com