Last updated: July 29, 2026
We don't keep your parsed content. It's stored temporarily for up to 1 hour and automatically deleted — long enough to deliver it to your agent, then it's gone.
We don't log what you parse. URLs you submit and the markdown we return are never persisted in logs.
We don't have accounts. There is no user database, no profile, no email collection unless you write to us first.
We don't track you. No cookies, no analytics scripts, no fingerprinting. The only thing we look at is your IP address, and only to stop abuse.
We don't sell anything about you. We sell a service (web-to-markdown parsing). We do not sell, rent, share, or monetize user data — ever.
NanoParse is an agent-native web-to-markdown API at nanoparse.app. You send us a URL. We render the page (JavaScript and all), extract the meaningful content, and return it as clean Markdown. Your AI agent pays $0.0175 USDC on Base per successful request via the open x402 protocol. That's the transaction. Nothing else.
Because the rendering and extraction take time, we use a queue-based architecture. Your request is accepted, processed asynchronously, and the result is held in temporary R2 storage for up to 1 hour — then it's automatically and permanently deleted.
Request URLs. URLs submitted for extraction are processed dynamically; rendered Markdown results are cached temporarily in R2 (1 hour TTL) for performance. Raw HTML is not retained after extraction.
IP Addresses. Used solely for abuse prevention — rate limiting and free-tier tracking. Stored in-memory per Worker isolate (typically 5-15 minute lifetime). Not logged to disk or external analytics services.
Payment Information. x402 payments occur on-chain (Base L2). We verify transaction hashes but never hold or custody funds. We do not have access to your wallet's private key.
Parsed markdown results are stored in R2 with a 1-hour Time-To-Live (TTL). They are automatically deleted after the TTL expires. Results are also deleted immediately upon successful retrieval by the requesting agent.
IP-based rate limit counters are stored in-memory per Worker isolate and in KV with short TTLs (typically 60-300 seconds). They are not permanently logged.
NanoParse uses Cloudflare Workers, R2, Queues, and Browser Rendering for infrastructure. Cloudflare's privacy policy governs their handling of network-level data. We use rate limiting for abuse prevention on the demo page. No other third-party services are used for data processing, analytics, or tracking.
We process the minimum data needed to deliver the service: URLs, IP addresses for rate limiting, and on-chain payment verification. No personal data is stored persistently. If you have questions about data handling or wish to exercise GDPR rights, contact hello@nanoparse.app. Since we don't maintain user accounts, identity verification for data requests will require proof of IP address ownership at the time of the relevant request.
We'll update this policy as the service evolves. Changes are effective immediately upon posting. If you continue using NanoParse after changes are posted, you accept the updated policy.
For privacy questions: hello@nanoparse.app.