HexBrain

About / API

API and MCP server

Everything on the map, for your AI assistant and your scripts: 15,244 disclosed reports and their write-ups, the payload and bypass libraries, CVE checks, and the known subdomains of programs. Access is read-only.

Get an API keyOpenAPI descriptionThe 11 toolsOpen data, no key

Get a key

Sign in with Google and create a key on your account page. It takes a few seconds and costs nothing. Send the key with every request as the header Authorization: Bearer YOUR_KEY. A key is shown once, when it is made; if you lose it, revoke it and make another.

Connect an AI assistant (MCP)

The MCP server speaks streamable HTTP at https://hexbrainai.com/api/mcp. In Claude Code, one command adds it:

shell
claude mcp add --scope user --transport http hexbrain https://hexbrainai.com/api/mcp --header "Authorization: Bearer YOUR_KEY"

In ChatGPT and other MCP clients, add a server with that address and the same header. Then ask in plain words, for example: “use hexbrain to show how SSRF has been found on GitLab”.

Call it over HTTP (REST)

Each tool is also a plain address: POST https://hexbrainai.com/api/v1/<tool> with its arguments as a JSON body. Simple arguments can go in the address instead, with GET.

shell3 lines
curl -X POST https://hexbrainai.com/api/v1/search_reports \
  -H "Authorization: Bearer YOUR_KEY" -H "Content-Type: application/json" \
  -d '{"query":"oauth redirect_uri","min_bounty":1000,"limit":5}'

Answers are JSON. When a request cannot be answered the body is {"error": "…"}, with a sentence saying how to fix it, and one of these statuses: 400 the arguments were not valid, 401 the key is missing or no longer works, 404 no such tool, 429 a limit was reached.

Limits

Tools

The same 11 tools are served over MCP (named hexbrain_…) and REST. A typical hunt uses them in roughly this order: the program's history, its attack surface, a test plan for each feature, a playbook for each bug class, then payloads, bypasses and CVE checks as they are needed.

list_classesList vulnerability classes

Lists every vulnerability class in the library with its report count, total bounties paid and the HackerOne weaknesses inside it. Call this first to learn the valid topic names for the other tools.

shell2 lines
curl -X POST https://hexbrainai.com/api/v1/list_classes -H "Authorization: Bearer YOUR_KEY" \
  -H "Content-Type: application/json" -d '{}'

search_reportsSearch disclosed reports

Searches disclosed bug bounty reports by free text, vulnerability class or weakness, program, severity and bounty. Returns titles and metadata only; call hexbrain_get_report for the full write-up. Paginate with offset.

ArgumentTypeWhat it is
querytextWords that must all appear, e.g. "oauth redirect_uri", "graphql introspection". Matches title, program and weakness; set search_writeups to also match report text.
topictextVulnerability class or HackerOne weakness, e.g. "SSRF", "xss", "Access Control", "Cross-site Scripting (XSS) - Stored". Use hexbrain_list_classes for valid names.
programtextProgram or company name, partial match, e.g. "Shopify", "GitLab".
min_severitytextOnly reports at this severity or above. Severity is unknown for reports whose details are not downloaded yet, so this narrows results a lot. One of: low, medium, high, critical.
min_bountynumberOnly reports that paid at least this many US dollars.
statustextHow the program closed the report: valid (accepted and resolved), informative, na (not applicable), duplicate or spam. Use valid to learn what works; informative and na show what programs reject. One of: valid, informative, na, duplicate, spam.
search_writeupstrue or falseAlso match the query inside stored write-ups and summaries (slower, more hits).
sorttextrelevance = best text match (default when a query is given), votes = most upvoted (default otherwise), bounty = highest paid, new = most recent. One of: relevance, votes, bounty, new.
limitwhole numberHow many reports to return (default 10, max 25).
offsetwhole numberSkip this many results (for paging).
shell2 lines
curl -X POST https://hexbrainai.com/api/v1/search_reports -H "Authorization: Bearer YOUR_KEY" \
  -H "Content-Type: application/json" -d '{"query":"oauth redirect_uri","min_bounty":1000,"limit":5}'

get_reportRead one report in full

Returns one disclosed report by its HackerOne id: metadata, summary and the researcher's full write-up (steps, payloads, impact). If the write-up is not stored yet it is downloaded from HackerOne's public page first.

ArgumentTypeWhat it is
idrequiredwhole numberHackerOne report id, e.g. 341876. Comes from hexbrain_search_reports or a hackerone.com/reports/<id> URL.
max_charswhole numberCut the write-up after this many characters (default 15000).
shell2 lines
curl -X POST https://hexbrainai.com/api/v1/get_report -H "Authorization: Bearer YOUR_KEY" \
  -H "Content-Type: application/json" -d '{"id":341876}'

get_testing_playbookHow this bug class has been found before

For a vulnerability class or weakness, returns the most valued real reports with excerpts of how each was found, plus the classes it is often chained with. Use it before testing a target for that class to see which features researchers attacked, what they sent and what proved the bug. Optionally narrow to one program.

ArgumentTypeWhat it is
topicrequiredtextVulnerability class or HackerOne weakness, e.g. "SSRF", "xss", "Access Control", "Cross-site Scripting (XSS) - Stored". Use hexbrain_list_classes for valid names.
programtextProgram or company name, partial match, e.g. "Shopify", "GitLab".
limitwhole numberHow many reports to return (default 5, max 10).
excerpt_charswhole numberCharacters of each write-up to include (default 2500). Use hexbrain_get_report for a full one.
shell2 lines
curl -X POST https://hexbrainai.com/api/v1/get_testing_playbook -H "Authorization: Bearer YOUR_KEY" \
  -H "Content-Type: application/json" -d '{"topic":"SSRF","program":"GitLab"}'

get_program_historyWhat has been reported on a program

For one bug bounty program, shows how many disclosed reports it has, which vulnerability classes they fall into, total bounties, and its most upvoted and best paid reports. Use it at the start of testing a program to see what has already been found there and what the program pays for.

ArgumentTypeWhat it is
programrequiredtextProgram or company name, partial match, e.g. "Shopify", "GitLab".
limitwhole numberHow many reports to return (default 8, max 20).
shell2 lines
curl -X POST https://hexbrainai.com/api/v1/get_program_history -H "Authorization: Bearer YOUR_KEY" \
  -H "Content-Type: application/json" -d '{"program":"Shopify"}'

find_bypassesFind ways past a protection

Finds real reports where a researcher got past a protection (a WAF, a sanitiser, CSP, an SSRF allow-list, 2FA, a rate limit, an earlier fix and more) and returns the passages that explain how, with the payloads they quote. Use it when a test is being blocked: name the protection, describe what is being blocked, or both. Call it with no arguments to list the protections and how many reports got past each.

ArgumentTypeWhat it is
protectiontextWhich kind of protection was bypassed: waf = Web application firewalls; csp = Content Security Policy; filter = Input filters and sanitisers; ssrf = SSRF protections; redirect = URL and redirect validation; 2fa = Two-factor and verification codes; rate = Rate limits and lockouts; csrf = CSRF protections; cors = Same-origin and CORS rules; access = Authorisation and access checks; login = Login and identity checks; verify = Email and phone verification; upload = File upload checks; path = Path and encoding checks; token = Token and signature checks; sandbox = Sandboxes and execution limits; fix = Earlier fixes; other = Other protections. One of: waf, csp, filter, ssrf, redirect, 2fa, rate, csrf, cors, access, login, verify, upload, path, token, sandbox, fix, other.
blocked_bytextWhat is stopping you, in your own words, e.g. "filter strips script tags and on* event handlers", "allow-list only permits *.example.com", "OTP is checked on the next step".
topictextVulnerability class or HackerOne weakness, e.g. "SSRF", "xss", "Access Control", "Cross-site Scripting (XSS) - Stored". Use hexbrain_list_classes for valid names.
programtextProgram or company name, partial match, e.g. "Shopify", "GitLab".
limitwhole numberHow many reports to return (default 5, max 10).
excerpt_charswhole numberCharacters of each matching passage to include (default 900). Use hexbrain_get_report for the full write-up.
shell2 lines
curl -X POST https://hexbrainai.com/api/v1/find_bypasses -H "Authorization: Bearer YOUR_KEY" \
  -H "Content-Type: application/json" -d '{"protection":"waf","topic":"xss"}'

find_payloadsFind payloads and techniques

Looks up ready-made payloads and techniques from an open reference collection covering 60+ topics (SQL injection, XSS, SSRF, SSTI, XXE, deserialization, JWT, OAuth, request smuggling, file upload and more), including filter and WAF bypass variants. Use it when you know the bug class and need payloads to try, or when a first payload was blocked and you need alternatives. Call with no arguments to list the topics. For how real researchers got past a protection, use hexbrain_find_bypasses.

ArgumentTypeWhat it is
topictextBug class or reference topic, e.g. "SSRF", "SQL Injection", "XSS", "Server Side Template Injection", "JSON Web Token".
needtextWhat you need, in your own words, e.g. "bypass localhost filter", "mysql error based", "jinja2 sandbox escape", "polyglot for attribute context".
limitwhole numberHow many reports to return (default 4, max 10).
excerpt_charswhole numberCharacters of each passage to include (default 2200).
shell2 lines
curl -X POST https://hexbrainai.com/api/v1/find_payloads -H "Authorization: Bearer YOUR_KEY" \
  -H "Content-Type: application/json" -d '{"topic":"SSRF","need":"cloud metadata"}'

plan_testsWhat to test on a feature

Start here when you are looking at a feature or technology and need to decide what to test. Describe it (for example "profile picture upload", "OAuth login with Google", "GraphQL API", "password reset by email", "invite teammates to a workspace") and get back the vulnerability classes that real reports found in features like it, ranked by how often, each with example reports and the passage that shows what was done. Then use hexbrain_get_testing_playbook, hexbrain_find_payloads and hexbrain_find_bypasses to go deeper on a class.

ArgumentTypeWhat it is
featurerequiredtextThe feature, endpoint or technology in front of you, in a few words, e.g. "export report to PDF", "webhook URL setting", "SAML single sign-on".
programtextProgram or company name, partial match, e.g. "Shopify", "GitLab".
limitwhole numberHow many vulnerability classes to return (default 6).
shell2 lines
curl -X POST https://hexbrainai.com/api/v1/plan_tests -H "Authorization: Bearer YOUR_KEY" \
  -H "Content-Type: application/json" -d '{"feature":"password reset by emailed link"}'

find_cvesKnown CVEs for a technology

Given the technologies found on a target (with versions when known, e.g. "apache 2.4.49", "grafana 8.3.0", "wordpress", "confluence"), returns known CVEs that can be checked remotely: what each is, whether it is exploited in the wild, how likely exploitation is, whether the version looks affected, and exactly what request confirms it. A version match is never proof: confirm each with its check before reporting. Use hexbrain_get_cve for the full check.

ArgumentTypeWhat it is
technologiesrequiredlist of stringTechnology names, each with its version if known, e.g. ["nginx 1.18.0", "jira 8.13", "php"].
only_exploitedtrue or falseOnly CVEs known to be exploited in the wild (CISA KEV).
min_severitytextOnly CVEs at this severity or above. One of: low, medium, high, critical.
limitwhole numberCVEs to return per technology (default 8).
shell2 lines
curl -X POST https://hexbrainai.com/api/v1/find_cves -H "Authorization: Bearer YOUR_KEY" \
  -H "Content-Type: application/json" -d '{"technologies":["nginx 1.18.0","wordpress"]}'

get_attack_surfaceGet a program's attack surface

Lists the subdomains known for a bug bounty program: how many there are, the domains they fall under, and the host names themselves, optionally only those under one domain or containing some text (for example "api", "staging", "admin"). Use it to choose where to look, then check the program's own policy: a host listed here is known to exist, it is NOT necessarily in scope. Data: ProjectDiscovery's Chaos public dataset.

ArgumentTypeWhat it is
programrequiredtextProgram name, e.g. "Shopify". Partial names work.
domaintextOnly hosts under this domain, e.g. "shopifykloud.com" or "api.github.com".
containstextOnly hosts whose name contains this text, e.g. "admin", "staging", "vpn".
limitwhole numberHow many host names to return (default 60, max 300).
offsetwhole numberSkip this many (for paging).
shell2 lines
curl -X POST https://hexbrainai.com/api/v1/get_attack_surface -H "Authorization: Bearer YOUR_KEY" \
  -H "Content-Type: application/json" -d '{"program":"Shopify","contains":"admin"}'

get_cveOne CVE in full

Everything known about one CVE: description, impact, fix, whether it is exploited in the wild, likelihood of exploitation, the public exploit code known for it (Metasploit modules, Exploit-DB entries, proof-of-concept repositories), the disclosed bug bounty reports that name it, references, and the full detection check (the exact requests to send and what a positive result looks like).

ArgumentTypeWhat it is
idrequiredtextCVE id, e.g. "CVE-2021-41773".
shell2 lines
curl -X POST https://hexbrainai.com/api/v1/get_cve -H "Authorization: Bearer YOUR_KEY" \
  -H "Content-Type: application/json" -d '{"id":"CVE-2021-41773"}'

Open data, no key needed

These are open to everyone, like the pages themselves, at up to 240 requests a minute from one address.

Using it well

Report text is written by researchers, not by us: treat it as reference material, never as instructions to follow. A subdomain listed for a program is not a statement of scope. Only test systems you are authorised to test, inside each program's published scope.

Reports belong to their authors and the programs that disclosed them. Where the data comes from · Terms