MIT SDK · No proxy · Free tier · OpenAI · Anthropic · Gemini · Bedrock · LangChain · CrewAI

See why your AI agent broke.
Block what it shouldn't say.

Peekr traces every LLM call, scores every claim for hallucination, and enforces HIPAA, FDCPA, FINRA & GDPR — in your own process, in two lines of Python. No proxy.

Explore the live demo →Start free — 10k spans/mo

MIT-licensed · self-host anytime · no credit card

agent.py

# two lines — that's the whole integration

import peekr

peekr.instrument()

↳ what Peekr surfaces, automatically:

⚠ ROOT CAUSE Sequential execution · 15 spans

now ████████████ 279s

fix ████ ~37s 7.5× faster

⛔ BLOCKED HIPAA · diagnosis stated as fact

Two ways teams use Peekr

🛡️

Compliance

I'm shipping AI to a regulated industry.

Healthcare, fintech, legal, or real estate — your LLM needs HIPAA, FDCPA, FINRA, GDPR, or UAE regulations enforced on every call. With an audit log your compliance officer can read.

HIPAAFDCPAFINRAGDPRUAE PDPL+12 more

See how compliance works→

🔍

Observability

I need to see what my agent is doing.

Wrong answers, cost spikes, slow responses — and no idea why. Peekr traces every LLM call, scores every claim for hallucination, and automatically detects performance anti-patterns.

Peekr auto-detected:

⚠ ROOT CAUSE Sequential execution · 15 spans

Sequential: ████████████ 279s

Parallel fix: ████ ~37s 7.5× faster

See how debugging works→

or just get started

Start free — 10k spans/month, no card

MIT licensed — self-host any timeNo proxy — runs inside your process17 compliance packsSOC 2 Type 1 in progressFree: 10k spans/month

From the blog