Hands-on testing of LLM features, RAG, agents, MCP tools and phone agents, before your customer's security review. Fixes arrive as pull requests your engineers review and merge.
Scanners are good at the standard web checklist. The risk in an AI product sits in chains: injected text, an agent decision, a tool call, another customer's data. That is the part we test by hand.
From $6,500AI agents
Your agent can call tools. We find out who else can make it.
Hands-on testing of LLM features, RAG, agents and MCP tools. We follow injected instructions all the way to a tool call and check what leaks between your customers.
Learn more
From $8,500Voice agents
We call your agent and test what a caller can make it do.
We phone your AI agent, try to break it, and check the parts a simulated caller cannot see: recordings, tool authorization, telephony and consent.
Learn more
From $6,500SaaS review
Proof that one customer can never see another.
A multi-tenant security review for Next.js, Supabase and Firebase apps. Every confirmed finding ships with a pull request your engineers review and merge.
Learn more
After the test: SOC 2 readiness and a fractional security lead for the same team.
Everything in writing, with dates. You always know what we test and what we must never touch.
/ 01
You answer six questions in writing and we send a written quote within one business day. No discovery call is required, though we are happy to have one.
/ 02
A certified tester, named in your proposal, works through your AI features, agents, phone agents and tenant boundaries by hand. Scanners cover the standard list so human time goes to chains and business logic.
/ 03
The report arrives within three business days of the last testing day, together with fix pull requests for confirmed findings.
A security test is only worth what your customer's reviewer accepts. We write every finding so it can be reproduced, fixed and retested.
Not a list of scanner output. A report written to be sampled, fixes your engineers can merge, and a letter you can send.
Fixes as pull requests close findings fast. These rules keep the evidence independent, and they are written into every scope.
Starting prices are on the pricing page. Your price is agreed in writing before we start.
Automated pentest tools
$1,500 to $4,000
Fast and cheap. Strong on the standard web list, weak on agent chains and business logic.
Wardkey
$6,500 to $12,000
A certified tester named in your proposal, fixes as pull requests, retest and a signed letter.
Large human-led firms
$10,000 to $16,000+
Deep, but quoted, slower to start and rarely write your fixes.
Every test is led by a certified tester. You see the name and certification in your proposal, before you sign.
The lead tester is named on the report and signs the attestation letter on behalf of Wardkey LLC.
One lead tester stays with your test from the first day to the retest. Fixes are retested by a tester who did not write them.
Our team's previous work. Client names stay private, the numbers come from the engagement record.
A production platform where businesses run AI phone and chat agents for their own customers, with telephony, payments and thirteen CRM integrations.
Read moreA B2B platform used by 340 business customers, tested with partial access (grey box).
Read moreAuditors look for scope tied to your systems, a named method, dated testing, findings with severity and evidence that fixes were retested. Our reports are written to that list.
We open pull requests from a fork or a protected branch, your engineers review and merge them, and a tester who did not write the fix retests. The attestation letter says who wrote each fix.
We start with staging or dedicated test tenants. Production is tested only with your written risk acceptance.
Wardkey LLC is a United States company. Testing is performed from the EU. We do not take work that requires US-person testers, such as FedRAMP, CJIS or ITAR scope.
Most tests start within two weeks of the signed scope and take four to eight tester days. The report follows within three business days.
Then the report says exactly what was tested and how, which is what your customer's reviewer needs to see.
Three defaults to check before your AI phone agent goes live: recordings that never expire, caller identity and tool authorization.
Read moreWhy a vector store shared by all your customers needs the same isolation tests as your database, and how we run them.
Read moreWe will tell you which lines a test answers, in writing, within one business day.