Wrap your Claude, GPT or Ollama calls in one line. Get EU AI Act audit-ready docs and real-time visibility into what your agents spend, how fast they respond, and where they fail — without any extra infrastructure.
Track spend, latency and errors per agent, per model, per day — all from the same logs you're already capturing for compliance.
Every log entry includes the actual USD cost based on current model pricing — claude-sonnet, gpt-4o, o3-mini and 15+ models supported out of the box.
Latency is captured automatically on every call. No extra instrumentation — the wrapper times itself.
Tag each client with agent_id="researcher". Get cost and error breakdowns per agent — see which one is eating your budget.
Cost spike (>2x yesterday), error rate >5%, p95 latency >10s — alerts fire automatically. No dashboards to configure.
from auditai import wrap_client, monitoring_stats import anthropic # Tag each agent for per-agent cost breakdown client = wrap_client( anthropic.Anthropic(), project="my-app", agent_id="researcher", # ← tag this agent session_id="conv-abc123", # ← group a conversation ) # Every call auto-logs: cost_usd, latency_ms, agent_id response = client.messages.create(...) # Get monitoring report for the last 24h logger = AuditLogger(project="my-app") stats = monitoring_stats(logger.read_all(), since_hours=24) # → total_cost_usd: $1.24 # → p95_latency_ms: 2340 # → per_agent: {"researcher": {"calls": 48, "cost_usd": 0.87}} # → alerts: [{"type": "cost_spike", "message": "Today 3.2x yesterday"}]
Already have auditai? pip install --upgrade auditai-sdk — monitoring is included from v0.2.5.
Install the SDK, answer 9 questions, get your compliance package — ready to send to any buyer's legal team.
Install — one line, no config
Works as a drop-in wrapper for Anthropic and OpenAI SDKs. Your existing code stays unchanged.
$ pip install auditai-sdk from auditai import wrap_client import anthropic # One line change — that's it client = wrap_client(anthropic.Anthropic(), project="my-product") # Every AI call is now logged and classified automatically
Classify your AI risk in 9 questions
Our wizard maps your use case to the EU AI Act risk categories: Minimal, Limited, High, or Unacceptable. Takes under 5 minutes.
$ auditai classify → Does your system affect credit decisions? [y/N] → Does it influence employment outcomes? [y/N] → Is it used in healthcare? [y/N] ... Classification: LIMITED RISK (Art. 52) Obligations: transparency notice, interaction logging
Generate your Article 26 Compliance Package
One command produces a PDF with everything an enterprise buyer or regulator wants to see: risk category, obligations checklist, technical evidence from your logs, and a deployer declaration.
$ auditai report --company "Acme SL" --email "cto@acme.com" ✓ Risk classification: LIMITED RISK (Art. 52) ✓ Obligations: 3 applicable, 3 covered by auditai ✓ Evidence: 14,832 calls logged, 0 anomalies ✓ Deployer declaration: signed Report saved: EU_AI_Act_Report_Acme_2026-05-07.pdf → Send this to your buyer's legal team
Enterprise buyers don't just ask for EU AI Act docs. They ask for SOC 2 Type II evidence too. auditai covers both — automatically.
One SDK integration. Two compliance frameworks. Zero extra work. Get your package →
Start free. Add what you need.
SDK
Open source. Wrap your calls in one line.
EU AI Act Templates
The 5 documents the law requires — ready to fill in.
Express Report
You fill a form. We generate the Art. 26 report PDF for you.
AI Monitor
Real-time cost, latency and error tracking for your AI agents.
Questions? marc@auditaisdk.com
Tell Marc what you need. He'll get back to you within 24 hours with a plan.
All 5 documents with sector-specific examples, risks and metrics. Ready to review and sign.
Healthcare AI
Annex III §5(a) — radiology, triage, clinical decision support. Dual compliance: EU AI Act + MDR 2017/745.
€79
Get Healthcare pack →HR & Recruitment AI
Annex III §4 — CV screening, candidate ranking, performance management. Fairness metrics included.
€79
Get HR pack →Fintech & Credit AI
Annex III §5(b) — credit scoring, SME lending. DORA + EBA guidelines. Gini, PSI, disparate impact metrics.
€79
Get Fintech pack →Education AI
Annex III §3(b) — student assessment, admission, at-risk prediction. GDPR Art. 8 (minors) included.
€79
Get Education pack →Justice & Law Enforcement AI
Annex III §8 — recidivism risk, judicial support. Art. 5 prohibited practices analysis included.
€79
Get Justice pack →Critical Infrastructure AI
Annex III §2 — energy grids, water, transport. NIS2 + IEC 62443 timelines integrated.
€79
Get Infrastructure pack →For law firms & consultancies
Reseller License
Use all 6 packs with unlimited clients. White-label allowed. Includes all future updates.
€800 one-time
Get reseller license →Stay up to date
Annual Updates
The law evolves. Get updated templates every time the EU AI Act guidance changes.
€19 /year
Subscribe to updates →Need a different industry? Ask Marc