Shaping agentic AI
Agentic AI Foundation
Vstorm was the first agentic AI engineering company to join the Agentic AI Foundation and received its 2026 Open Agentic AI Innovation Award.
Read the announcementAI agent audit and due diligence
An engineering verdict before you scale an agent or sign a deal
A clear path to the agent hitting its target, or a clear no before you invest. Run on your real cases: what breaks, what it costs at volume and what to fix first.
50+ workflows agentified since 2017
-
~20% → ~40% conversion to a final saleThree-agent product advisor guiding customers through print-order configuration.
Read case study -
42% → 98% raw LLM vs guideline-executing accuracy on the open 50-scenario benchmarkHIPAA-compliant guideline-executing triage — 42% to 98% on the open 50-scenario benchmark.
Read case study -
Up to 1 hr → ~2 min to create a workflowText-to-workflow agents building validated node graphs inside the platform.
Read case study -
Supply chain intelligence agents reducing manual coordination overhead.
-
AI agent implementation for a global automotive enterprise.
-
Every inquiry answered in English and Arabic, from ARIJ's own knowledge baseBilingual English/Arabic agent inside ARIJ's Moodle environment, answering only from ARIJ's own knowledge base.
Why agent reviews miss the risk
The risk sits in the cases nobody demoed.
A demo hides data access, test gaps, cost at volume and model or vendor ties.
- 01
Where it stallsThe review is a demo and slides.
What we do We read the code and run unprepared cases.
- 02
Where it stallsAccuracy quoted from the team's own tests.
What we do We check what the evaluations leave out.
- 03
Where it stallsRunning cost guessed from the pilot invoice.
What we do Cost per task, projected to planned volume.
How the review runs
A ranked list of risks you can act on
Engineers read and run the system; every finding cites its evidence.
- 0103
Scoped to the decision
We write down the decision the report serves: scale as is, fix first, or price the risk.
- Review questions agreed in writing
- Code and log access
- A delivery date in writing
- 0203
Every layer read and run
Engineers rerun the claimed results on fresh cases, check test gaps and measure cost per task.
- Architecture and code review
- Data access and tool security
- Model and vendor dependency
We hold our own builds to the same test. On an open 50-scenario benchmark a raw LLM scored 42% and the agent we built scored 98%. - 0303
Findings ranked, fixes planned
Owners get fixes in order of risk; investors see what must be fixed before closing.
- Findings with evidence and estimates
- Remediation plan ranked by risk
- Evaluation cases you can rerun
Engineering in the open
We help build what agentic AI runs on
Open standards, open-source tooling and published research, all public, so you can check our engineering before you hire us.
Open source
We build open-source agent tooling that other engineering teams install
10M+ PyPI downloads
4,000+ GitHub stars
33 public repositories
Framework contributors
When the framework breaks, we fix it at the source
We maintain our own open-source agent framework and contribute fixes to the frameworks our agents run on, through official partnerships with their teams. Breaking changes reach us before they reach your codebase.
Research
Our engineers and PhDs publish peer-reviewed research
Peer-reviewed papers on agent architectures, evaluation and AI security, and an open benchmark anyone can rerun against our results.
Track record
Frontier AI teams now build on what we shipped first
Our client agents were in production before most of today's agent frameworks had a stable release. You get a team that already knows what breaks between a promising demo and a system people rely on every day.
Case studies
Agent systems we built and measured
Engineering workflows built in minutes, validated before they run
The agent drafts each workflow and checks it in several steps before it runs, so engineers skip the manual setup.
Up to 1 hrto prepare a workflow by hand
~2 minto create one with the agent
A print-order advisor that lifted conversion to a final sale
Three agents and 15 tools walk customers through print-order configuration, across more than a billion product combinations.
~20% → ~40%conversion to a final sale
96%+web-to-print orders handled end to end
Clinical triage that executes the guideline, step by step
The model answers the assessment questions. The guideline decides the disposition, the same way every time.
42% → 98%accuracy, raw LLM vs guideline-executing
96%weighted F1 across 566 scenarios
Engagement models
Ways to start
Each path starts with a free call with a founder or senior AI engineer. Changes ship only once they meet the agreed target.
Start with the agent you run
We measure your agent on your own cases and agree the target before changing it.
Best whenYour agent runs, but accuracy, cost or speed falls short.
FAQ
Questions about AI agent audits
What does AI agent audit and due diligence include?
A review of seven areas, each with its own risk rating: architecture, code quality, evaluation coverage, data handling and access, cost per task, vendor and model dependency, and security risks. Engineers read the code, run the system on cases nobody prepared and project cost per task to the planned volume. You receive a written report with evidence, a rating, a fix and an estimate for every finding, plus the evaluation cases to rerun after the fixes.
Who is an AI agent audit for?
Two kinds of buyer. Teams that run an agent system and want an outside engineering view before they scale it, and investors or acquirers who need to know what they are buying before a deal closes. Both get the same review; the report answers different questions.
Can you review a system built by another vendor or with AI coding tools?
Yes. The review starts from the code and the running system as they are, whoever wrote them. For vibe-coded systems we pay extra attention to tests, error handling and secrets, which are often missing.
What access do you need?
Read access to the repository, the infrastructure configuration and the logs, plus a sample of real cases to run the system on. For a deal review we agree the form of access with both sides before the work starts.
How long does an AI agent audit take?
There is no fixed length: we agree the scope and delivery date in writing before the review starts, and for a deal we plan back from the closing date. The size of the system and how quickly access is granted set the date.
What do we receive at the end?
A written report listing every finding with its evidence, a risk rating and the recommended fix with an estimate, ranked by risk. You also keep the evaluation cases built during the review, so the same checks can be rerun once the fixes are in.
Does the audit include a security test?
It reviews security risks at the design and code level: prompt injection paths, tool permissions, credential handling and what data the agent can send out. It is not a penetration test. If you need one, the findings tell the testers where to start.
Will you fix what the audit finds?
If you want us to. The remediation plan is written so that your own team, or any other engineering team, can follow it. Fixes run as a separate engagement, usually through performance optimization.
Before you scale or invest
Get an engineering view of an agent system
Scope and delivery date agreed in writing; for a deal, planned from closing.
From the blog
Reading before a review
Agentic AI
AI agents can act: can you still control them?
Agents that issue refunds need identity, authorisation, approval and audit before the call lands. What an agentic AI control plane is, and when you need one.
Agentic AI
Lesson 3: System access is a hard constraint
Why system access (VPNs, API keys, ERP exports) belongs on the critical path of an agentic AI project, with a named owner and deadline.
Agentic AI
AI proof of concept vs production-grade agent: key differences is design and intent
A proof of concept shows an agent can do the work. A production agent has to keep doing it. The design and intent differences between the two.