agentprime
  • Ex-Datadog security leadership
  • Fortune 500 security programs
  • Fixed-price engagements

Security services for software companies

Your product grew faster than your security program.

Senior security expertise for software companies without a security team. We assess the product you already run, including the code your AI tools wrote, get you ready for the audits and questionnaires your customers send, and stay on as your security partner.

You work directly with Craig and Justin.No account managers, no junior hand-offs.

Ex-DatadogSecurity product leadership
Fortune 500Security programs designed and built
8+ yearsRunning a SaaS company that buys this work
Fixed priceScope and fee agreed in writing first

Sound familiar?

Security became somebody's second job.

Nobody at a 20-person company wakes up wanting to run a security program. It shows up anyway: in a questionnaire, an insurance renewal, a customer's security team on a sales call. In plain words, it sounds like this.

  • A big prospect sent a 200-question security questionnaire. The deal's stuck until we answer it.

    deal blocked
  • Our compliance dashboard is all green. Nobody has ever actually tried to break in.

    compliance != security
  • Half our codebase was written with Cursor. I honestly don't know what it skipped.

    AI-written code
  • Security is my second job. I do it at 11pm, when I do it at all.

    no owner
  • We have a pen test vendor, a scanner and a consultant. Nobody owns the whole picture.

    vendor sprawl
  • Our insurer asked about MFA, backups and access reviews. I guessed.

    insurance renewal

What we keep finding

AI made shipping faster. It made the old mistakes faster too.

The same handful of problems shows up in almost every growing product we look at, and AI coding tools make them more common. We find them, fix them with you, and leave a check in place so they don't come back.

F-01high

One customer can see another's data

The most common serious bug we find, and AI coding tools make it more common. The page checks you’re logged in, but not that the record is yours. Change an ID in the URL and you’re reading someone else’s invoices.

FixScope every query to the signed-in account, and test it with two accounts.

F-02high

Keys with far more power than they need

A full-access payment or messaging key where a read-only one would do, sitting in an env file on three laptops and one AI agent’s context. One leak and someone else runs your account.

FixRestricted keys per service, kept out of code, rotated after anyone leaves.

F-03medium

Nothing stops a bot from running up the bill

No caps on SMS, email or AI model spend. A stolen key or a scripted signup form can send 50,000 texts on your account overnight, and you find out from the invoice.

FixDaily caps and alerts on every paid API, plus rate limits on public endpoints.

F-04medium

No trail when something goes wrong

Logging only covers the part of the app someone was debugging last month. When a customer asks what happened to their account, there’s nothing to look at.

FixUser ID and IP on every request, logs kept 90 days, alerts on login spikes.

security-review · PR #482Blocked
  • ✓Secrets in code0 found
  • ✓Dependencies0 critical
  • ✗Authorization1 high
GET /api/invoices/:id returns any account's invoice. Customer A can read customer B's billing history by changing the ID.
Fix: scope the query to the signed-in account invoices.ts:41
  • ✓Rate limits on paid APIspresent
  • ✓Request logginguser + IP
Ran on every pull requestDeploy held until fixed

How to work with us

Three ways to engage

Most companies start with an assessment, then keep us on as their security partner. Every engagement has a fixed scope and a fixed price.

02 · 21 days

Audit-Ready Sprint

Three weeks to get a growing software company ready for SOC 2, enterprise security reviews and harder questions from insurers.

  • SOC 2 gap analysis
  • AI and agent access review
  • Team access and laptop review
  • A 90-day fix plan and questionnaire answers
Everything included →
03 · Monthly

Ongoing Coverage

A security lead on call each month, for teams that keep shipping after the review is done.

  • Expert review of high-risk changes
  • Help with enterprise security reviews
  • Monthly alert and log review
  • Team access check-ins
Everything included →

Why not the usual options

A badge isn't security.

Compliance software collects evidence. Big consultancies bring big-company process and hourly bills. Neither was built for a small team shipping AI-written code every week.

Compliance tool aloneBig consultancyagentprime
Someone actually tries to break itNo. It tracks controls.Yes, at enterprise pricesYes
Reads the code your AI tools wroteNoRarelyYes. It is where we start.
Who does the workYou do, with a dashboardA partner sells it, juniors deliver itThe senior people you met
PriceSubscription, plus add-onsHourly, and the scope growsFixed. The fee is a cap.
Built for a 5-50 person teamYesNoYes
After the reportMore dashboardA new statement of workA check on every deploy, and us on call

How an assessment works

From first call to a clear plan in about three weeks.

  1. Week 0

    30-minute call

    What you run, what's coming up and what's worrying you. Free.

  2. Week 1

    Scope and access

    Fixed scope and price in writing. You grant read-only access.

  3. Weeks 1-2

    The assessment

    Threat model, code, cloud, identity, AI tools, limits and logging.

  4. Week 3

    Report and walkthrough

    Findings ranked by risk, a 90-day roadmap and a one-page summary.

  5. After

    Stay covered

    We retest your fixes, and the check in your pipeline keeps running.

What you walk away with

Not a 90-page PDF nobody reads.

  • 01

    Findings ranked by risk

    Each one with where it is, why it matters and the fix, written so your team or your AI tools can apply it.

  • 02

    A roadmap and a one-page summary

    A 90-day plan with owners, and a summary for your board, your biggest customer or your insurer.

  • 03

    A check that keeps running

    A security review step in your deploy pipeline, tuned to your stack, so next month's code gets checked too.

Security summary

Example product · Security Assessment

2 high open
Critical0
High2
Medium5
Low9
Fix first
  • HighScope invoice and profile queries to the signed-in account
  • HighRemove admin access for two former contractors
  • MedAdd daily caps and alerts on SMS and email sends

How we work

Built to be easy to trust.

You're giving us a look inside your business. These are the rules we hold ourselves to, and they're in the contract.

  • 01

    Fixed scope, fixed price

    Agreed in writing before we start. The fee is a cap. No surprise invoices.

  • 02

    Read-only by default

    We use each service's own access controls. We never ask for passwords or API keys.

  • 03

    Your systems, your call

    We never change production without your written approval.

  • 04

    Your findings stay yours

    Everything we learn about your systems is your confidential information.

  • 05

    AI with care

    We use AI to move faster, only under terms that keep your data out of model training.

  • 06

    Plain language

    Ranked findings your engineers can fix, and a one-page summary for everyone else.

Who we are

A founder who's been the buyer. A security leader who's built programs for the Fortune 500.

Enterprise-grade judgment, sized for a 20-person company. You work with us directly, not a junior team.

More about us →

Craig Hewitt

Co-founder, client lead

Founder and CEO of Castos. Has been the buyer of security work, and runs every engagement from the client's side of the table.

Justin Massey

Co-founder, security lead

Led a security product at Datadog. Has designed and built security programs for Fortune 500 companies, from the product side and as a consultant. Leads the technical work on every engagement.

Questions

Straight answers

Something else on your mind? Email us.

Is this a pen test?

Not by itself. A pen test checks whether someone can break in from outside at one moment in time. We review the code, the configuration, who has access and how you ship, and we leave a check running on every deploy. If a customer or auditor needs a formal pen test, we arrange one with vetted testers and manage it for you.

Do you do SOC 2?

We get you ready for it. We don’t issue SOC 2 reports, because we aren’t auditors. The Audit-Ready Sprint shows you the gaps an auditor will find, gives you a plan to close them, and works alongside whichever auditor or compliance platform you pick.

We built most of our product with AI. Is that a problem?

No. That’s exactly who we work with. AI coding tools let small teams ship real products fast. They also repeat a predictable set of mistakes, and we know where to look for them.

We already use a compliance platform. Do we still need this?

Probably, yes. Compliance platforms collect evidence and track whether controls are in place. They don’t try to break your app, read the code your AI tools wrote, or tell you which of 200 controls actually matter for a team your size. We work alongside whichever platform you use, and the fixes we recommend count toward its checklist.

Will you need our passwords or API keys?

No. We use each service’s own read-only access wherever possible. Never send passwords, API keys or tokens by email, chat or document. If one arrives anyway, we delete it and ask you to rotate it.

Do you fix what you find?

We give you specific fixes, ranked by risk, written so your team or your AI tools can apply them. We don’t write or deploy your production code, and we never change your systems without your written approval. We can set up checks and guardrails with you, and you review them before they run.

What size companies do you work with?

Mostly software companies with roughly 5 to 50 people and $1M to $5M in annual revenue, especially AI-first teams and companies in regulated industries like finance and healthcare. Most already have a product in production and customers asking harder questions. We also review new products before launch.

How does pricing work?

Every engagement has a fixed scope and a fixed price, agreed in writing before we start. The fee is a cap: we won’t bill more for that scope. Ongoing Coverage is a flat monthly fee with included hours, and you can cancel with 30 days’ notice.

We think we've been breached. Can you help?

Incident response isn’t part of these services. If you’re in the middle of an incident, email us anyway and we’ll point you to someone who handles it. Once things are stable, a review is a good way to make sure it doesn’t happen again.

Not sure where you stand? Let's find out.

Tell us what you run and what's coming up: a big customer, a security questionnaire, an insurance renewal. We'll tell you plainly whether we can help and what it would take.

You work directly with Craig and Justin.No account managers, no junior hand-offs.

Talk with a founder →