Enterprise procurement asking for EU AI Act proof?
Show them one structured package instead of scattered internal docs.
EU AI Evidence Builder
Turn scattered AI evidence into one review-ready EU AI Act package
Your team already has the material — draft text, logs, monitoring outputs, risk notes, and approvals. EUAgentAI helps you shape it into one structured EU AI Act package with evidence links, open blockers, and exportable HTML / JSON / PDF outputs.
Choose the route first: full EU package or standalone Article 50. Then complete the Builder gates that determine the package run.
Self-hostedOpen-source coreYour runtime data stays with you
Is this you?
Do any of these sound familiar?
EU client or partner requesting compliance evidence?
Assemble review-ready documentation they can actually work with.
Investor due diligence flagging regulatory risk?
Demonstrate structured compliance readiness before it becomes a blocker.
AI system needs Article 50 transparency marking?
Build a standalone transparency package with the evidence you already have.
Two paths
Choose your package path
Choose the route first. Full-package work continues through Builder scope gates. Standalone Article 50 starts directly in the Article 50 builder flow.
Full EU package
High-risk AI documentation package
For teams preparing the broader high-risk documentation package. Start in the Builder, then confirm role, scope, and whether Article 50 is in or out.
- Annex IV + related statutory articles
- Builder-gated provider / deployer route
- Article 50 can be included or excluded in scope
- One structured package for review or handoff
Route: Builder -> role and scope gates -> package run
Standalone Article 50
Transparency evidence package
For systems where transparency is the only package you need right now. This starts directly in the standalone Article 50 builder flow.
- Focused on Article 50 transparency duties only
- Separate builder and evidence path
- Starts directly in the Article 50-only route
- Available as standalone €199 package help
Route: Standalone Article 50 builder -> transparency package run
Public proof
Public sample packages and reference runs
Start with our controlled sample package. The cards after it are generated from the published proof surfaces, so new live reference runs appear from config and published bundles instead of manual page edits.
Controlled sample
Sample EU package
Our controlled demo of the minimum EU package, published first so you can inspect the baseline output shape before running the workflow on your own system.
- Controlled demo package
- Readable EU report
- Article-level outputs
GitHub reference run
Skene external pilot
Real self-hosted acceptance run against the public Skene repository, used as a clean external GitHub reference bundle for the EU package path.
- Real GitHub repository run
- External reference workflow archetype
- Handoff-ready reviewer package + EU outputs
GitHub reference run
QClaw external pilot
Real self-hosted acceptance run against the public QClaw repository, showing a support and operations agent runtime packaged into a handoff-ready EU bundle.
- Real GitHub repository run
- Support and operations agent archetype
- Handoff-ready reviewer package + EU outputs
GitHub reference run
Ragpi external pilot
Real self-hosted acceptance run against the public Ragpi repository, showing a documentation-backed AI assistant packaged into a handoff-ready EU bundle.
- Real GitHub repository run
- Docs and knowledge assistant archetype
- Handoff-ready reviewer package + EU outputs
The problem
Why teams get stuck before the first package
The blocker is rarely one missing paragraph. It is the package never getting into one review-ready version.
Evidence is scattered across 5+ tools
Draft text lives in Google Docs, logs in Datadog, test results in CI, risk notes in Notion, declarations in email threads.
Nobody owns the last mile
Engineering, legal, and compliance each hold part of the evidence. Nobody has the mandate or the template to assemble it.
Reviews send it back
Reviewers cannot work with a folder of raw files. They need one structured package with cross-references and clear status.
Every new request restarts from scratch
Without a reusable package structure, every customer audit or procurement request means rebuilding the same evidence.
Outputs
What you get
One package your reviewer can actually use:
One structured package instead of scattered files
Go from draft text, logs, and internal records to one organized package your team can review and hand over.
Reach the first real package faster
Use one builder, one structure, and one handoff path instead of rebuilding evidence in documents and chat threads.
See what is still blocking handoff
Track the records, references, and approvals that still need to be completed before review or submission.
Give reviewers less to untangle
Hand legal, compliance, governance, or customer review one structured package instead of a folder of raw files.
Honest scope
What this does — and does not do
What it does
- Assembles scattered evidence into one structured package
- Provides Annex IV documentation templates
- Generates review-ready HTML, JSON, and PDF exports
- Links technical evidence to statutory sections
- Self-hosted — your data stays in your environment
What it does not do
- Replace legal counsel or risk classification
- Automate conformity assessment decisions
- Provide legal advice or regulatory interpretation
- Guarantee compliance — final sign-off stays with your team
Best for
- Small AI teams (30–250 people) with 1–6 high-risk systems
- Teams that already have evidence but need to assemble it
- Companies facing procurement, customer, or investor review
- AI agent vendors selling to EU or regulated customers
Pricing preview
FREE
Free self-serve EU AI Act path
Free
Use the open-source repo on your own. No commercial support is included.
- Unlimited self-serve evaluation
- EU starter + open-source docs
- Builder + templates
- Live demos
- No setup help included
ARTICLE 50
Article 50 transparency help
EUR199
Use this when Article 50 is your only EU scope. This usually applies if your system talks to people as AI, generates synthetic text, image, audio, or video, uses emotion recognition or biometric categorisation, or produces deepfakes or public-interest synthetic content.
- 1 standalone Article 50 package
- Article 50 scope check
- Help to reach the first real transparency package
- Short results review + next-step handoff
- Focused help for Article 50 evidence
PAID HELP
First package help for one real system
EUR499
Use this when your team needs hands-on help getting to the first real package on its own system.
- 1 supported agent
- Adapter fit check
- Help to reach the first real package
- Short results review + next-step handoff
- Focused help for the first package
FAQ
Frequently asked questions
What problem does the EU AI Act Package Builder solve?
It helps teams turn scattered draft text, logs, test outputs, declarations, monitoring notes, and internal records into one organized EU AI Act package for a specific AI system.
Why do teams struggle to prepare an EU AI Act package?
The work usually stalls because system scope, intended use, risk records, technical evidence, legal wording, and approvals sit in different tools. The Builder gives the team one package structure and one handoff path instead of rebuilding the same material in documents, tickets, and folders.
What documents are included in the EU AI Act package?
The package can include a readable dossier, linked technical materials, article-level outputs, evidence indexes, monitoring and logging support, declaration-related records, and the human-owned sections that still need review or sign-off. The exact package depends on the selected role, scope, and product path.
Which EU AI Act articles and annexes does this product cover?
The full package path is built around high-risk AI system documentation and evidence, including Annex IV technical documentation and article-level outputs such as risk management, data governance, logging, transparency, human oversight, accuracy and robustness, quality management, monitoring, serious incident support, conformity assessment, declaration, marking, and registration records where they are in scope.
Who is this for?
It is for AI teams, compliance owners, governance teams, legal teams, and operators who need to prepare a review-ready EU AI Act package for one AI system without losing the technical evidence behind the written dossier.
When should a team use the full EU package?
Use the full EU package when the system needs the broader high-risk AI documentation path, including Annex IV-style technical documentation, risk and oversight records, monitoring support, and the related article-level outputs for review.
When should a team use the standalone Article 50 package?
Use the standalone Article 50 package when transparency is the package you need right now: for example when the system talks to people as AI, generates synthetic text, image, audio, or video, uses emotion recognition or biometric categorisation, or produces deepfakes or public-interest synthetic content.
What is the difference between the starter and the full package?
The starter is a first self-serve check on your own running agent or existing report directory. The full package is the broader review package: it combines selected role and scope, written dossier sections, linked evidence, exports, and the handoff material your team needs to complete and approve.
What happens after we start the builder?
You choose the package path, fill the role and scope gates, add the human-owned context, attach or generate supporting materials, review the package, and then either continue self-serve or send the package context for paid first-package help.
Do we need to share our AI model or runtime data?
No. The core workflow is designed for self-hosted use. Runtime evidence, generated reports, and reviewer outputs can stay in your controlled environment.
Can this produce both human-readable documents and technical evidence?
Yes. The goal is to keep readable review documents and technical supporting materials in the same package so legal, governance, engineering, and compliance reviewers are not working from disconnected files.
Does this replace legal review?
No. This is a package builder and evidence workflow. Legal interpretation, classification decisions, residual-risk judgment, and final sign-off stay with your team or counsel.
Ready to assemble your first package?
Start with the free builder or see a sample package first.