HUMAN Security launches open-source MCP server for AI threat analysis

On July 24, 2025, HUMAN Safety introduced the open-source launch of its Mannequin Context Protocol Server, making a bridge between AI assistants and the corporate’s menace intelligence infrastructure. The launch represents a shift towards conversational safety interfaces, permitting safety groups to question advanced information by way of pure language moderately than conventional dashboards.

The MCP Server connects AI assistants like Claude and Cursor on to HUMAN’s menace detection APIs. Safety analysts can now ask questions in plain English and obtain solutions backed by menace intelligence that displays over 20 trillion interactions per week globally. This strategy eliminates the necessity to navigate a number of interfaces or write customized queries to entry safety information.

The announcement described the system as enabling safety groups to work together with menace information “in essentially the most human approach doable: by speaking to it.” The server processes queries in real-time, pulling info from HUMAN’s Sightline Cyberfraud Protection and Shopper-side Protection merchandise. Outcomes embody assault traits, person safety profiles, cluster forensics, and compliance standing.

Technical implementation and necessities

The MCP Server operates regionally on safety staff infrastructure. As soon as configured with HUMAN API credentials, the server interprets queries between AI assistants and HUMAN’s APIs. The system requires an energetic HUMAN account, legitimate API credentials, an MCP-compatible AI shopper, and Node Package deal Supervisor put in on the person’s laptop.

The structure depends on the Mannequin Context Protocol, developed by Anthropic. MCP gives standardized connections between AI programs and information sources, enabling giant language fashions to entry exterior APIs in a managed method. OpenAI CEO Sam Altman introduced in March 2025 that OpenAI would add help for Anthropic’s MCP throughout its merchandise, indicating rising trade adoption of the protocol.

HUMAN’s implementation permits safety groups to question particular assault sorts, examine suspicious accounts, overview cluster standing, and assess visitors safety posture by way of conversational prompts. The system can generate visualizations based mostly on queries, reworking textual content requests into dashboard shows exhibiting menace summaries from specified time intervals.

Cyberfraud protection capabilities

The server gives entry to assault development evaluation throughout a number of dimensions. Safety groups can request breakdowns by assault sort, together with account takeover, scraping, and abuse. Time-series information reveals energetic campaigns, focused endpoints, and protection responses. The system tracks sophistication scores, bot signatures, spoofing habits, IP fame, and assault cluster info.

Account investigations ship complete safety profiles for particular person customers. Queries return threat scores, habits anomalies, incident timelines, recognized triggers, and transaction patterns. The server can analyze total visitors safety posture in real-time, calculating ratios between reputable and blocked requests, assault quantity traits, net versus cellular threats, and safety management effectiveness.

In keeping with the Affiliation of Nationwide Advertisers’ Q1 2025 Programmatic Transparency Benchmark, solely 41% of programmatic advert spend resulted in high quality impressions. This implies practically 60% of spend was misplaced to non-quality stock, representing an estimated $21.6 billion in unrealized media worth globally. HUMAN’s conversational safety strategy addresses these inefficiencies by lowering time-to-insight for fraud detection.

Shopper-side safety integration

The MCP Server connects to HUMAN’s Shopper-side Protection product, enabling queries about third-party scripts working on fee pages. Safety groups can request full lists of scripts by web page, together with vendor attribution, threat ranges, PCI DSS compliance standing, and recognized vulnerabilities with CVE identifiers. This visibility helps organizations perceive their provide chain by way of single questions.

Compliance auditing capabilities analyze safety headers in opposition to PCI DSS necessities. The system evaluates content material safety insurance policies, HTTP Strict Transport Safety settings, and different header configurations. Outcomes embody actionable fixes mapped to particular compliance necessities. Safety groups can question latest client-side incidents to view DOM injections, script anomalies, and cross-site scripting makes an attempt with threat scores, timestamps, and affected URLs.

Vendor threat evaluation gives ranked assessments based mostly on habits, privileges, recognized vulnerabilities, and entry patterns. The server identifies which distributors pose the best threat to checkout flows, serving to safety groups prioritize mitigation efforts. HUMAN’s dashboard gives the great view that safety professionals can now question conversationally.

OpenAI integration and allowlisting

OpenAI documented HUMAN’s integration as a part of its ChatGPT agent allowlisting system. The documentation explains that ChatGPT agent indicators each outbound HTTP request utilizing the HTTP Message Signatures normal, enabling organizations to confirm genuine visitors. HUMAN seems in two allowlisting configurations for the ChatGPT agent.

HUMAN Sightline categorizes ChatGPT agent as a trusted AI agent in its Recognized Bots & Crawlers system. Organizations utilizing Sightline or Bot Defender can allow the agent by navigating to insurance policies, deciding on visitors coverage settings, trying to find ChatGPT Agent, and altering the rule to permit. No customized signature verification is required as a result of HUMAN performs this validation.

AgenticTrust, HUMAN’s newer product, gives cryptographic verification of ChatGPT agent exercise. The system displays intent in each session and permits the agent to learn, log in, and make purchases by default. Organizations can modify these permissions by accessing insurance policies, trying to find ChatGPT Agent, and granting or revoking particular permissions as wanted.

The mixing between HUMAN and OpenAI displays broader trade coordination round AI agent safety. ChatGPT’s instant checkout capabilities launched on September 29, 2025, enabling direct purchases by way of conversational interfaces. This commerce performance requires strong safety infrastructure to stop fraud whereas sustaining person expertise.

Trade context and agentic commerce

The MCP Server launch happens as organizations grapple with AI agent governance. In keeping with HUMAN Safety’s annual cybersecurity report launched in 2024, 80% of corporations utilizing HUMAN’s platform selected to dam recognized giant language mannequin user-agents. Considerations included mental property theft, content material scraping, and AI-enhanced cyber assaults. This widespread blocking demonstrated the stress between AI adoption and safety necessities.

AgenticTrust addresses these considerations by offering visibility into AI agent habits. The product surfaces, classifies, and governs AI brokers with adaptive belief mechanisms. Organizations can enable useful automation whereas stopping abuse and defending enterprise operations with out sacrificing development. Google Cloud’s survey of three,466 executives in 2025 revealed that 88% of early adopter organizations implementing AI brokers reported optimistic return on funding.

The conversational safety mannequin aligns with predictions that AI brokers will develop into mainstream enterprise infrastructure. McKinsey analysis estimated that generative AI’s impression on productiveness may add trillions of {dollars} in worth to the worldwide financial system, with AI brokers enjoying a essential function. The financial impression depends upon organizations safely integrating AI capabilities into operations.

HUMAN’s strategy differs from shopper AI instruments by specializing in enterprise safety necessities. MIT analysis analyzing 300 AI initiatives recognized a basic distinction between instruments like ChatGPT that improve particular person productiveness and enterprise implementations affecting revenue and loss efficiency. The MCP Server targets operational safety moderately than common productiveness.

Product updates and roadmap

The Q3 2025 product replace doc detailed a number of enhancements past the MCP Server. HUMAN Sightline Cyberfraud Protection added direct Auth0 actions, essential well being alerts, and improved person search performance. The detection modifications audit web page gives visibility into updates from the earlier seven days, highlighting new energetic detectors, updates to current detectors, deletions, and enablement standing modifications.

AgenticTrust established itself as an innovation in agentic commerce safety throughout Q3 2025. The product empowers clients with full visibility and governance over shopper AI brokers. Organizations can detect and classify AI brokers, confirm belief ranges, and govern how brokers work together with net functions. The system evaluates intent behind every agent-initiated interplay.

Shopper-side Protection added aggregated PCI DSS dashboard views, webhook integrations, and API listing updates. Clients can now view their PCI DSS dashboard and stock throughout a number of software IDs and host domains concurrently. The webhook integration permits self-configurable notifications throughout channels. Organizations can obtain alerts for client-side incidents on checkout pages, PCI DSS notifications for scripts and headers, and configurable abstract studies.

HUMAN’s promoting safety platform enhanced compliance capabilities and integration well being instruments. The compliance dashboard added fields together with stock accomplice area, proprietor area, vendor relationship, first advert system, and vendor identifiers. Month-to-month utilization studies grew to become out there for MediaGuard and FraudSensor, permitting clients to obtain proactive notifications about request volumes.

Safety implementation concerns

The MCP Server requires organizations to take care of safety greatest practices whereas gaining conversational capabilities. The server runs regionally, making certain that menace intelligence queries stay inside organizational infrastructure. API credentials present entry controls, limiting which customers can question particular information units. The structure separates the conversational interface from the underlying safety intelligence.

Organizations should confirm that intermediate proxies protect signature headers, together with Signature, Signature-Enter, and Signature-Agent headers. These headers allow cryptographic verification of AI agent requests. HUMAN performs verification robotically when organizations use its merchandise, eliminating the necessity for customized signature checking code.

The system addresses basic challenges in menace detection operations. Safety groups usually face friction when interacting with menace information by way of a number of dashboards, handbook queries, and time-consuming triage processes. Fashionable threats transfer rapidly, requiring intelligence programs that transfer sooner. Conversational interfaces cut back time-to-insight by eliminating context switching between instruments.

Context engineering emerged as a vital self-discipline for AI agent success in 2025. Trade practitioners recognized info structure because the figuring out issue for AI implementations. HUMAN’s MCP Server applies context engineering rules by structuring menace intelligence for optimum AI consumption. The system determines acceptable info for particular duties, codecs information for mannequin processing, and adapts context based mostly on question necessities.

Market implications for safety groups

The conversational safety mannequin modifications how organizations workers and practice safety operations. Conventional safety analyst roles required experience in a number of dashboard interfaces, question languages, and information interpretation methods. Conversational interfaces decrease the barrier to accessing menace intelligence, probably enabling broader staff members to contribute to safety operations.

HUMAN clients embody Fortune 500 corporations and main platforms worldwide. The corporate examines greater than 20 trillion interactions per week globally, offering in depth visibility into menace patterns. This scale permits the MCP Server to ship solutions backed by complete information moderately than restricted samples.

The promoting trade faces specific challenges from invalid visitors and fraud. HUMAN and Opera Ads announced a partnership in September 2024 to combine pre-bid detection into Opera Adverts’ platform. Related TV promoting presents distinctive vulnerabilities on account of restricted on-device measurement capabilities and heavy reliance on server communication. HUMAN’s fraud protection resolution addresses these challenges by way of behavioral evaluation moderately than static detection strategies.

Trade research proceed revealing important programmatic spending waste. Hidden intermediaries, duplicative auctions, and fraud inside advanced provide paths drain advertiser budgets. Common publishers work with over 10 supply-side platforms, which means the identical impression typically auctions a number of occasions to 1 purchaser. This synthetic competitors inflates prices per thousand impressions by way of waste moderately than real demand.

Implementation timeline and availability

HUMAN launched the MCP Server as open-source software program beneath MIT license on July 24, 2025. The corporate printed the code to its GitHub repository, enabling clients and safety researchers to look at the implementation. Organizations can start utilizing the server instantly by putting in it from the repository and configuring it with legitimate HUMAN API credentials.

The announcement indicated that HUMAN plans to develop the MCP toolset and help extra workflows over time. The corporate dedicated to releasing tutorials, use instances, and deeper integration steering in subsequent updates. Buyer Success groups present setup help and data to current HUMAN clients.

Organizations with out HUMAN accounts can request demonstrations or contact gross sales to judge the platform. The MCP Server requires energetic HUMAN accounts with API entry, limiting speedy availability to present clients. This strategy ensures that menace intelligence entry stays managed whereas the conversational interface expands capabilities.

The server works with MCP-compatible AI shoppers together with Cursor and Claude. As extra AI assistants undertake the Mannequin Context Protocol, the server will help these platforms with out requiring code modifications. This standardization aligns with trade actions towards interoperable AI infrastructure.

HUMAN described the discharge as “only the start” of conversational safety capabilities. The corporate indicated ongoing growth to boost tooling and develop supported workflows. Safety groups can anticipate extra options that construct on the conversational question basis.

Timeline

Abstract

Who: HUMAN Safety, a cybersecurity firm defending Fortune 500 corporations and main platforms, launched the MCP Server. Authors Ben Diamant and Ori Gold introduced the product. Safety groups utilizing HUMAN’s Sightline Cyberfraud Protection and Shopper-side Protection merchandise can implement the server.

What: The HUMAN Safety MCP Server is an open-source bridge between AI assistants and HUMAN’s menace intelligence APIs. The server permits conversational queries about assault traits, person safety profiles, cluster forensics, visitors safety posture, third-party script dangers, compliance standing, and client-side incidents. The system processes queries in real-time utilizing the Mannequin Context Protocol.

When: HUMAN introduced the MCP Server on July 24, 2025. The product grew to become instantly out there to current HUMAN clients by way of the corporate’s GitHub repository beneath MIT license. Implementation requires API credentials from energetic HUMAN accounts.

The place: The MCP Server runs regionally on organizational infrastructure, connecting to HUMAN’s cloud-based menace intelligence platform. The system operates globally, serving organizations that use HUMAN’s safety merchandise throughout a number of areas and industries.

Why: Safety groups face friction interacting with menace information by way of dashboards, queries, and handbook triage. The conversational interface reduces time-to-insight by enabling pure language queries as a substitute of studying a number of interfaces. HUMAN goals to make safety intelligence accessible by way of essentially the most intuitive methodology: dialog. The discharge addresses trade wants as organizations implement AI brokers whereas sustaining safety necessities.