canonical: https://jentic.com/apis/googleapis.com/webrisk # Google Web Risk API The Google Web Risk API lets client applications check URLs against Google's continuously updated lists of unsafe web resources, including phishing pages, malware sites, and unwanted software. It exposes both a hash-based lookup that preserves user privacy and a direct URI search for low-volume use, alongside an evaluate endpoint that returns risk scores for borderline URLs. Web Risk is a paid commercial successor to the public Safe Browsing API and is intended for products that need higher quotas, an SLA, and the ability to submit URLs for review. ## For AI agents Check URLs against Google's threat lists for phishing, malware, and unwanted software, and submit suspicious URLs for review. Designed for moderation, link-shortening, and email security pipelines. ## Scope Does not handle vulnerability scanning of your own apps, bot detection, content moderation of text or images, or DNS filtering - use for URL threat lookups against Google's hosted threat lists only. ## Capabilities - Check a URL for phishing, malware, or unwanted-software classifications using uris:search - Compute privacy-preserving hash prefix diffs against threat lists with threatLists:computeDiff - Verify candidate hash prefixes locally and confirm full hashes via hashes:search - Submit a suspicious URL to Google for review through the submissions endpoint - Track long-running submission and evaluation operations with the operations endpoints ## Use cases ### Link safety check before redirect URL shorteners, email gateways, and chat platforms call uris:search before resolving a user-supplied link to block known phishing and malware destinations. The endpoint accepts a URL and a list of threat types and returns matched threats. Because Web Risk is paid and quota-managed, it suits high-traffic products that have outgrown the free Safe Browsing API. Example prompt: Call GET /v1/uris:search with the URL https://example.com/login and threatTypes=MALWARE,SOCIAL_ENGINEERING and report whether any threat is returned. ### Privacy-preserving local threat database Browsers and security clients sync a local copy of Google's threat lists by calling threatLists:computeDiff and only requesting full hashes from hashes:search when a candidate prefix matches. This pattern keeps actual user URLs on the client and sends only truncated hashes upstream. Web Risk's diff format is identical in shape to Safe Browsing v4 so existing client logic ports cleanly. Example prompt: Call GET /v1/threatLists:computeDiff with threatType=MALWARE and a stored versionToken, then update the local hash prefix database with the returned additions and removals. ### Submitting newly discovered phishing URLs Security teams that detect phishing pages targeting their brand submit those URLs to Google through POST /v1/{+parent}/submissions so that the entry can be added to threat lists and protect the wider Chrome and Android ecosystem. Submissions are processed as long-running operations; the operations endpoints track status. This is a useful integration for fraud and trust-and-safety workflows. Example prompt: POST a submission with uri=https://fake-bank.example.com to /v1/projects/{project}/submissions and poll the returned operation until done is true. ### Agent-driven URL screening through Jentic Moderation and customer-support agents call Web Risk through Jentic to vet links before sending them to users or following them autonomously. The agent queries Jentic for check a url for phishing, loads the uris:search schema, and executes with the URL pulled from the conversation. Credential isolation in Jentic means OAuth tokens for the underlying Google project never enter the model's context. Example prompt: Through Jentic, call GET /v1/uris:search with a URL extracted from a user message and threatTypes=MALWARE,SOCIAL_ENGINEERING,UNWANTED_SOFTWARE, then refuse to forward the link if any threat is returned. ## Key endpoints | Method | Path | Description | | --- | --- | --- | | GET | /v1/uris:search | Check a URL against threat lists | | GET | /v1/hashes:search | Resolve hash prefixes to full hashes | | GET | /v1/threatLists:computeDiff | Compute a diff of a threat list | | POST | /v1/{+parent}/submissions | Submit a URL for review | | GET | /v1/{+name}/operations | List long-running operations | | POST | /v1/{+name}:cancel | Cancel a long-running operation | ## Key resources - **uris** — Search a single URL against threat lists - **hashes** — Resolve hash prefixes to full hashes for client-side verification - **threatLists** — Compute incremental diffs of the local hash database - **submissions** — Submit URLs to Google for review - **operations** — Track long-running submission and evaluation operations ## Why Jentic - **Setup:** Wiring the Web Risk API by hand means setting up OAuth 2.0 client credentials, managing short-lived access tokens, and matching lookup and submission calls against the webrisk.googleapis.com host. Through Jentic you install once, import the Web Risk API from the API Directory, store the OAuth credential once, and your agent calls it. - **Permission scoping:** The Web Risk API carries its submission target in the request body and query, so you limit the agent to the operations it needs, such as searching URIs or hashes against Google's threat lists. You choose the operations it may call, so submitting new threat reports is not included unless you add it. - **Credential handling:** Your Web Risk OAuth credential is stored once, encrypted, by your own Jentic One instance and injected at execution time. Its refresh token never enters the agent's prompt, logs, or context. - **Discovery method:** Agents search Jentic by intent such as 'check a URL for phishing' or 'look up a hash against threat lists', and Jentic returns the uris:search or submissions operation with its input schema so the agent calls the right endpoint without browsing the reference docs. ## Related APIs - **Safe Browsing API** — Free, lower-quota sibling that exposes the same threat lists without an SLA - **reCAPTCHA Enterprise API** — Detect abusive traffic and bots before a URL is even followed - **Web Security Scanner API** — Scans your own App Engine and Compute apps for vulnerabilities ## FAQ ### What authentication does the Web Risk API use? Web Risk uses OAuth 2.0 (the Oauth2 and Oauth2c security schemes in the spec) scoped to https://www.googleapis.com/auth/cloud-platform. Through Jentic the OAuth credentials are stored in the encrypted vault and exchanged for short-lived access tokens at execution time. ### How is Web Risk different from the free Safe Browsing API? Web Risk is the paid commercial product with higher quotas, an SLA, and support for URL submissions through POST /v1/{+parent}/submissions. Safe Browsing v4 is free but rate limited and does not include the submission endpoint. The hash diff and lookup logic is otherwise compatible. ### What are the rate limits for the Web Risk API? Web Risk quotas are configured per Google Cloud project in the Cloud Console under APIs and Services and depend on your billing plan. Google publishes the contractual QPS in the Web Risk pricing documentation; for hash-based lookups the local cache built from threatLists:computeDiff handles the bulk of traffic so live calls stay well under quota. ### How do I check a URL for phishing through Jentic? Search Jentic for check a url for phishing, load the schema for GET /v1/uris:search, and execute it with the uri parameter and threatTypes including SOCIAL_ENGINEERING and MALWARE. Jentic returns the matched threats array; an empty response means the URL is not currently on a list. ### Can I submit phishing URLs that target my brand? Yes. POST /v1/{+parent}/submissions accepts a URI and creates a long-running operation. Track it through the operations endpoints (GET /v1/{+name}/operations or POST /v1/{+name}:cancel) until the operation completes. Approved submissions are added to Google's threat lists and propagate to Chrome and Android. ### Can I limit what my agent is allowed to do with the Web Risk API? Yes. Because Jentic One is self-hosted, you run your own instance and your own rules decide which Web Risk operations and which OAuth credential the agent may use. You can allow only read-only threat lookups such as GET /v1/uris:search and GET /v1/hashes:search while excluding the submission endpoint (POST /v1/{+parent}/submissions), so the agent can check URLs against Google's threat lists but cannot file new threat reports unless you explicitly add that operation. The OAuth credential is stored by your instance and injected at execution time, keeping it out of the agent's context.