Auditbase gives every AI agent action a cryptographic fingerprint — tamper-proof, instantly verifiable, forever retrievable. One API call. Full accountability.
Modern AI agents don't just answer questions — they send emails, modify databases, execute transactions, call APIs, and make decisions on behalf of your users. When something goes wrong, can you answer: what did the agent do, when, and why?
Standard logs are mutable. Anyone with DB access can alter them. When a dispute arises or a regulator asks questions, you have nothing to prove your agent acted correctly.
When an agent behaves unexpectedly, you're left guessing. Which action caused the issue? What payload did it send? What was the exact sequence of events?
Finance, healthcare, and legal industries increasingly require AI systems to maintain auditable records. Building this yourself is weeks of work — and still fragile.
Every time your AI agent takes an action, Auditbase generates a SHA-256 hash of the exact event data — agent name, action type, payload, and timestamp. That hash is stored immutably alongside the event.
Later, you can call our verify endpoint to recompute the hash from stored fields and confirm the record hasn't been altered. If the hashes match, the log is untampered. If they don't, you know something changed.
curl -X POST https://api.auditbase.dev/v1/log \
-H "Authorization: Bearer sk_live_..." \
-H "Content-Type: application/json" \
-d '{
"agent": "billing-agent",
"action_type": "charge_customer",
"actor": "user_8a3f92",
"payload": {
"amount": 4900,
"currency": "usd",
"plan": "builder"
}
}'{
"id": "3f2a8b1c-9d4e-4f6a...",
"hash": "a9c3b8f2e1d4c6a7b9e2...",
"created_at": "2026-03-26T19:00:00Z",
"retained_until": "2027-03-26T19:00:00Z"
}Any team deploying AI agents that take real-world actions needs an audit trail.
Log every email sent, ticket updated, or refund issued by your support agent. Know exactly what your bot told customers — and when.
Every AI-triggered transaction, fraud flag, or account action gets a cryptographic receipt. Essential for PCI compliance and customer dispute resolution.
Prove your AI made decisions based on correct data at a specific point in time. Tamper-proof logs satisfy audit requirements without building custom infrastructure.
Track every file written, command executed, or API called by your coding agent. Debug regressions by replaying the exact sequence of actions that caused them.
Log every recommendation, triage decision, or alert sent by clinical AI. Meet HIPAA documentation standards without custom logging infrastructure.
Multi-agent pipelines touching CRMs, ERPs, and internal systems need a central audit record. Auditbase works across every agent in your stack.
Integrate in minutes. No infrastructure to manage. No schema to design.
Sign up, get a key in seconds. Start on the free tier — no credit card required.
POST to /v1/log with your agent name, action type, and any payload. One call.
Get back a SHA-256 cryptographic hash. This is your proof the event occurred.
Call /verify to recompute the hash and confirm the record is untampered.
A focused API that does one thing exceptionally well.
SHA-256 hashes computed from a canonical string of all log fields. Any modification to any field is immediately detectable via the verify endpoint.
Filter logs by agent, action type, actor, or date range. Cursor-based pagination for high-volume queries. Get exactly the events you need.
Store any JSON payload up to 256KB per log entry. Capture the full context of every agent action — inputs, outputs, model used, session ID, anything.
Real-time usage stats via /v1/usage. Track log volume by billing period. Rate limiting built in — 1,000 requests/minute per key.
Create and revoke API keys programmatically. Keys are stored as SHA-256 hashes — the raw key is returned exactly once and never persisted.
Logs retained 30 days on the free plan, up to 3 years on Studio and Scale. Automatic cleanup — no manual housekeeping required.
Any log entry can be verified at any time. Call the verify endpoint with a log ID and Auditbase recomputes the SHA-256 hash from stored fields, comparing it against the original.
If the hashes match — valid: true — the record is exactly as it was when the agent logged it. Nothing has been altered.
This is your receipting system for AI actions. Store the hash in your own systems if you want an additional check outside of Auditbase entirely.
{
"valid": true,
"computed_hash": "a9c3b8f2e1d4c6a7...",
"stored_hash": "a9c3b8f2e1d4c6a7...",
"verified_at": "2026-03-26T20:00:00Z"
}{
"valid": false,
"computed_hash": "a9c3b8f2e1d4c6a7...",
"stored_hash": "ff291abc94d1e820...",
"verified_at": "2026-03-26T20:01:00Z"
}Start free. Scale as your agents grow. No hidden fees, no per-seat nonsense.
Standard log systems (Datadog, CloudWatch, etc.) are great for observability — but they're not built for accountability. Anyone with access can modify or delete records. Auditbase computes a cryptographic hash at write time, so tampering is detectable. It's designed specifically for proving what an AI agent did, not just storing what happened.
The hash is computed from a canonical string built from: log ID, project ID, agent name, action type, the full serialized payload, and the ISO timestamp. Any change to any field produces a completely different hash, making tampering immediately detectable.
Yes. Auditbase is a simple REST API — it works with any language, any framework, and any AI agent. LangChain, CrewAI, AutoGPT, custom agents — if your code can make an HTTP request, it can log to Auditbase.
The API returns a 429 with a monthly_limit_exceeded error code. Your existing logs and hashes remain intact. Upgrade your plan anytime to restore logging — no data is deleted.
Not yet — we're a new product. Enterprise compliance certifications are on the roadmap. If compliance is a hard requirement today, reach out and we can discuss your specific needs.
Free tier available. No credit card required. One API call away.