Kaplaix Documentation
Kaplaix is an evidence-grade governance platform for AI agents. It captures every action your agents take, scores operational risk in real time, gates dangerous operations before they execute, and turns every session into auditable evidence.
What Kaplaix does
- Notary — Immutable, hash-chained event capture. Every tool call, data movement, approval, and environment change is recorded with cryptographic integrity.
- Risk Engine — Real-time risk scoring aligned to the OWASP Agentic Top 10. Ten dimensions, explainable factors, no black-box numbers.
- Guardian Mode — Pre-execution policy enforcement. The SDK checks with Kaplaix before the dangerous tool runs. If a policy says “requires review”, the agent waits for a human decision.
- Investigation Engine — Causal chain reconstruction across events. One-screen narrative of what happened and why.
- Evidence Packages — Signed ZIP exports (PDF + JSON + manifest + SHA-256) ready for auditors and regulators.
- Agent DNA — Behavioral fingerprinting per agent. Detects drift from established baselines using Hellinger distance.
Quick links
- Quickstart (5 minutes) — Send your first event
- TypeScript SDK — Full SDK reference
- Python SDK — Full SDK reference
- API Reference — REST API endpoints
- Event Schema — The 7 canonical event categories
Integration options
| Method | Effort | Best for |
|---|---|---|
| TypeScript SDK | 5 min | Node.js / Deno agents |
| Python SDK | 5 min | Python agents |
| LangChain adapter | 2 min | LangChain agents (TS + PY) |
| HTTP API | 10 min | Any language |
| Collector (coming soon) | 15 min | Log-based agents, no code changes |
Last updated on