12 AI agents that turn a short intake form into board-ready technology documents — architecture review, build-vs-buy decision, security posture, technical debt register, diligence pack, board brief. Each one drafts, reconciles and scores itself before you read it. Built for companies without a full-time CTO, and for the fractional CTOs and advisors who serve them.
Every agent is a named artefact, never a topic. Here is the job each group of them is built to finish.
Architecture Review, Build vs Buy and Scalability Plan take your stack as you describe it and return a written assessment: what holds, what breaks first, and what to do about it.
Engineering Org Design, Delivery Metrics and Technology Roadmap turn team shape and delivery pain into a structure, a measurement system, and a sequenced plan.
Security Posture Review, Technical Debt Register and Cloud Cost & Vendor Review write down the exposure — security, debt, spend — as a register you can prioritise and act on.
Technical Due Diligence Pack, Board Technology Brief and AI Adoption Strategy produce the documents you hand to a buyer, a board, or an investment committee.
These agents produce documents, not measured business outcomes. What you get is a structured, reconciled, rubric-scored deliverable built from the information you enter — the judgement about acting on it stays yours.
Not a replacement for engineering judgement — an accelerator that does the drafting, structuring and formatting, so the expensive hours go to the decisions.
Paste your own Anthropic API key and fill in a focused intake form — 15 fields describing your stack, team, constraints and the decision in front of you. Nothing plugs into your repos, cloud accounts or ticketing.
The agent drafts the document section by section, runs a pre-delivery QC pass that checks the sections agree with each other, and scores the result against a 100-point rubric. You pick the model before it runs.
Open the finished deliverable in the browser, download it as HTML or Excel, or print to PDF and take it into the board meeting.
You aren't buying a chatbot. You're buying the document — so that's what each agent hands you.
Every one of the 12 agents follows the same shape. This is the Technical Due Diligence Pack, section by section, exactly as the agent writes it.
Choose Advanced (Claude Opus 5) for the deepest reasoning, or Standard (Claude Sonnet 5) when you want a faster, cheaper pass on your own key. The model picker sits in the app; the price you pay Anthropic is whatever that model costs, with no markup from us.
Cardinal is the fictional company built into the suite as its worked example: a Series B freight-broker platform, a five-year-old Rails monolith on an end-of-life Ruby version, 38 engineers, an 800-million-row event table nobody has archived, a settlement module with no clear owner after two departures — and an acquisition conversation opening inside nine months.
Load Cardinal into the Technical Due Diligence Pack and the agent writes up the buyer's lens on that estate: what a diligence team would flag, what it would cost commercially, and what to fix first. Run the Technical Debt Register or Security Posture Review on the same company and you get consistent findings, because all 12 agents share the sample.
Mulkern AI Systems does not publish client case studies or customer outcome figures for this suite. When there is a real, permissioned result to cite, it will appear here and be identified as such.
License a single agent, one bundle, or the whole suite. Each agent is a live web app — open it, enter your key and licence, and generate.
Annual by default. Single and Bundle also come in 3- and 6-month terms for one-off engagements. Every license is bring-your-own-Anthropic-key.
A bundle always beats licensing its agents one by one, and the full suite is the right fit for a firm or a multi-client practice. Need more than 5 seats? Add them at $2,000/seat/year.
The Fractional CTO is one suite in a wider set. Mulkern AI Systems also builds the Fractional CFO, CEO, COO, CHRO, CRO and CMO suites on the same engine — same intake-to-deliverable model, same QC scoring, same bring-your-own-key economics. The Full Office licence covers every suite under one agreement rather than buying each one separately.
The Full Office is $63,818.18/yr — all 88 agents across the seven suites, five seats included (add more at +$2,000/seat). That is $14,181.82 less than licensing the seven suites separately ($78,000), and every future agent we add is included.
Anthropic API usage is billed directly to your own key — the cost depends on which model you choose for the run. License fees cover access to the agents only, with no markup on usage.
Minimum passing score: 85 / 100. Deliverables below the threshold are flagged before delivery so you can re-run rather than ship a weak document.
Each agent runs on your own Anthropic key, so your technical information is never stored on our servers and you control exactly what you spend. Our inference cost is zero — which is why the licence is priced the way it is.
There is no agent in your cloud account, no repo access, no read-only IAM role to approve. You describe the situation in an intake form and the agent writes from that — which is exactly what keeps security review short.
Every agent writes the same eight-section structure, runs a QC pass over its own output, and scores it before you see it. You get something you can print and take into a room, not a transcript to tidy up.
12 AI agents, each of which produces one board-ready technology document from a short intake form — architecture review, build-vs-buy decision, scalability plan, engineering org design, delivery metrics, technology roadmap, security posture review, technical debt register, cloud cost and vendor review, technical due diligence pack, board technology brief, and AI adoption strategy.
No. There are no integrations of any kind. Each agent is a web app with an intake form: you describe your stack, team and situation in 15 fields, and the agent writes the document from what you tell it. Nothing reads your code, your infrastructure or your tickets.
Your own Anthropic (Claude) API key and a licence. You paste the key into the app, fill the intake, choose your model, and generate. The key takes about two minutes to set up the first time.
You choose at runtime, in the app: Advanced (Claude Opus 5) for the deepest reasoning, or Standard (Claude Sonnet 5) for a faster and cheaper pass. Because it runs on your key, you pay Anthropic's rate for whichever you pick, with no markup from us.
Everything runs on your own Anthropic key. What you enter passes through our gateway only in the moment to call Claude — it is not written to disk, logged, or stored. We keep your licence details only. Since nothing connects to your systems, the only information involved is what you choose to type in.
Treat every deliverable as an expert-quality first draft, not a final decision. It is structured, self-reviewed and rubric-scored, so it is well past a raw chatbot answer — but a qualified engineer or advisor should review it before you act, especially on security and diligence findings. The agent only knows what you told it in the intake.
Not for this suite — it is new. We would rather show you the sample company built into every agent and let you judge the output than publish results we cannot stand behind. When there is a real customer result we have permission to share, it will be published and identified as one.
Every deliverable is shaped by your intake — your stack, constraints, team and the decision you're facing — and you can re-run as many times as you like to refine it. Suite owners can also have a custom agent built for a specific workflow.
Twelve documents a fractional CTO gets hired to produce, each finished in minutes on your own key.