Have I Been Pwned
Check email addresses, domains and passwords against the HIBP breach corpus from your AI agent. 5 tools work anonymously; 12 need an OAuth token and a paid plan. 17 tools, streamable HTTP.
Verified connector
Listed by Anthropic as a partner connector in its Connectors Directory.
Connection checked by Agentman on .
Anthropic states this reflects the level of review a connector received, not a security audit.
Connect Have I Been Pwned via MCP
https://haveibeenpwned.com/mcpWorks in any MCP-compatible client. In Agentman Studio it is one click — no config file to edit.
Have I Been Pwned Tools & Capabilities (17)
List public HIBP breaches, optionally filtered by domain, spam-list flag, and verification status.
Look up a single public HIBP breach by its canonical breach name, such as Adobe.
Return the most recently added public breach currently loaded into HIBP.
List the data classes used across public HIBP breach models, such as email addresses or passwords.
Query the public Pwned Passwords k-anonymity API with a 5-character SHA-1 or NTLM prefix and return matching suffixes with prevalence counts.
Search HIBP for breaches affecting a single email address. This tool requires an OAuth bearer token.
Query the authenticated HIBP k-anonymity breached-account range endpoint with the first 6 characters of a SHA-1 email hash.
Search for public pastes containing an email address. This tool requires an OAuth bearer token.
Return breached aliases for a verified domain. This tool requires an authorized subscription via OAuth bearer token.
List the domains associated with the authenticated HIBP subscription.
Return the current subscription details and feature flags for the authenticated OAuth-linked HIBP subscription.
Return website domains observed in stealer logs for an email address. Requires an authenticated subscription with stealer-log access.
Return email addresses observed in stealer logs for a website domain. Requires an authenticated subscription with stealer-log access.
Return email aliases and associated website domains observed in stealer logs for an email domain. Requires an authenticated subscription with stealer-log access.
Generate the TXT record value required to verify domain control via DNS. Requires an authenticated subscription with domain-verification access.
Complete domain verification by checking the expected HIBP TXT record on the target domain. Requires an authenticated subscription with domain-verification access.
Send a domain verification email to an approved alias such as admin or security. Requires an authenticated subscription with domain-verification access.
Read from the server on 2026-08-16, including each tool's own safety annotations.
Limits
- Most tools need a paid plan. Only 5 of 17 work anonymously. HIBP's documentation states that subscriber-only features follow the normal HIBP plan rules, and that a tool requiring a paid plan needs the signed-in account to hold a subscription including that capability.
- Domain data requires verified control. HIBP's API documentation states organisations must add a domain to their dashboard and verify control before searching it. There is no way to query an arbitrary third-party domain.
- Rate limits apply and vary by plan. HIBP's API documentation states limits depend on the key purchased and that exceeding them returns HTTP 429 with a retry-after header. Persistent excess may draw Cloudflare blocks or JavaScript challenges returning HTTP 503. Domain search has no formal limit but is intended for infrequent use; Pwned Passwords has none.
- Range results must be discarded. HIBP's terms, cited in its API documentation, require that k-anonymity results not matching the address searched are immediately discarded and not stored or further processed.
- Sensitive and retired breaches are held back. HIBP's API documentation states the public API returns no email addresses for a breach flagged sensitive, and that retired breach data is permanently removed and never returned. Email search — direct or by range — excludes both. An address can therefore be in a breach the connector will not report.
- Breach dates are approximate. HIBP's breach model documentation says the breach date is frequently long before discovery and should be used as a guide only, and that its
PwnCountis usually lower than media figures because of duplication in source data. - Unverified and fabricated breaches are in the corpus. The breach model carries
IsVerifiedandIsFabricatedflags, andhibp_list_breachesdefaults to including unverified breaches. A hit is not automatically a confirmed compromise. - No client secrets. HIBP's agent guide states confidential clients and client secrets are not supported, only public clients with PKCE — which constrains server-side deployments.
- No credential contents, ever. The connector reports that an address or password appeared in a breach. It cannot return what the password was.
- We did not exercise any tool. Our check was a read-only
initializeandtools/listhandshake. Behaviour above comes from the server's own tool descriptions and HIBP's documentation, not from calling anything.
Frequently asked questions
Do you need a Have I Been Pwned account to use the MCP server?
No, not for everything. Have I Been Pwned's documentation states that public breach metadata and data class tools work without signing in, and our anonymous handshake returned all 17 tool definitions on 2026-08-16. Searching an email address, a domain, pastes or stealer logs requires an OAuth bearer token for the hibp.mcp scope, tied to an active HIBP subscription.
Does the connector send my password to Have I Been Pwned?
No. The connector never sends or receives a password. HIBP's API documentation describes the Pwned Passwords k-anonymity model: your client hashes the password, sends only the first five hex characters, and receives roughly 800 hash suffixes with prevalence counts. Your client matches its own full hash locally, so HIBP never learns which password was checked.
Does a breach lookup return the leaked passwords or stolen data?
No. HIBP's breach model returns metadata only: a breach name, date, description and a DataClasses array naming the types of data exposed, such as email addresses or passwords. It names the categories, never the values. No tool in this connector returns a password, a credential or the contents of any breached record.
Which Have I Been Pwned tools can change anything?
Three of the seventeen. The server annotates hibp_generate_domain_verification_dns_token, hibp_verify_domain_verification_dns_token and hibp_send_domain_verification_email as readOnlyHint false with destructiveHint false. All three are domain-ownership steps, and the last one sends an email to an alias such as admin. The other fourteen tools are annotated read-only.
Why does a Have I Been Pwned tool return 403 or say it is unavailable?
Your HIBP subscription does not include that tool's tier. HIBP's troubleshooting notes say to confirm the signed-in account has an active subscription covering the feature being called, and that domain tools additionally require the domain to be verified in HIBP first. Its agent guide gives stealer logs as the example of a higher-plan feature.
Why are my Have I Been Pwned MCP requests being rate limited?
You are being rate limited, and the response tells you how long to wait. HIBP's agent guide states a 429 response includes retryAfterSeconds and rateLimitResetAt fields to honour before retrying. Its API documentation adds that limits depend on the purchased key, that domain search has no formal limit, and that Pwned Passwords is unlimited.
Why does the Have I Been Pwned connector return 401?
Your OAuth token is missing, expired or no longer valid for the HIBP MCP resource. HIBP's documented fix is to remove the saved credential from your client and reconnect so it obtains a fresh token. If authorisation will not complete at all, HIBP advises removing the connection first, since some clients cache failed registration or consent state.
Sources
- Have I Been Pwned MCP Server documentation (retrieved 2026-08-16; HTTP 200, no redirects). Source for the public/authorised split, setup steps, example prompts, the troubleshooting section and the per-tool read-only, open-world and destructive justifications. · retrieved 2026-08-16
- HIBP agent authentication guide (retrieved 2026-08-16; HTTP 200, no redirects). Source for the PKCE requirement, the no-client-secrets constraint, the scope list and the unsupported registration models. · retrieved 2026-08-16
- HIBP MCP server card (retrieved 2026-08-16; HTTP 200). Source for the transport declaration and the auth split summary. · retrieved 2026-08-16
- HIBP OAuth protected resource metadata (retrieved 2026-08-16; HTTP 200). Source for the
hibp.mcpscope and header-based bearer usage. · retrieved 2026-08-16 - HIBP API v3 documentation (retrieved 2026-08-16; HTTP 200, no redirects). Source for the k-anonymity contracts and prefix lengths, the padding behaviour, the breach model and its flags, domain verification requirements, rate limiting, and the discard-non-matching-results term. · retrieved 2026-08-16
- HIBP published agent skill (retrieved 2026-08-16; HTTP 200). Source for the per-tool public/authenticated grouping and the 401/403/429 error guidance. · retrieved 2026-08-16
- Live
initializeandtools/listhandshake againsthttps://haveibeenpwned.com/mcp— 17 tools with descriptions, input schemas and safety annotations, returned to an anonymous request (2026-08-16). Source for the annotation split, parameter details and the tool groupings. · retrieved 2026-08-16 - Anthropic Connectors Directory entry (snapshot taken 2026-08-16). Source for the partner tier and the 17 catalogued tool names. · retrieved 2026-08-16
- HIBP privacy policy — · Terms of use — · Pricing (all retrieved 2026-08-16; HTTP 200, no redirects). · retrieved 2026-08-16
Use in Agentman
Connect once and your agents call these tools on their own — on a schedule, in a workflow, with nobody at the keyboard.
Open in Agentman StudioServer Info
- Category
- Analytics
- Developer
- Have I Been Pwned
- Tools
- 17
- Domain
- haveibeenpwned.com
Using Claude Desktop or another MCP client? Setup docs — the connection URL above works anywhere.