PRODUCT

Safety infrastructure for
every AI agent.

Authensor wraps your agent's tool calls with policy evaluation, content scanning, and cryptographic audit trails. One integration point, full safety coverage.

ARCHITECTURE

Five layers of safety.

1

Agent Layer

Your agent calls guard() before every tool invocation. One function call, synchronous, zero config needed.

@authensor/sdk@authensor/langchain@authensor/openai
2

Policy Layer

Declarative rules define what actions are allowed, denied, or require human approval. Policies are versioned and auditable.

@authensor/engineBuilt-in policy (offline)Control plane policy (online)
3

Safety Layer

Content scanning runs in parallel with policy evaluation. Prompt injection, PII, credentials, and exfiltration patterns caught before they reach the outside world.

@authensor/aegis@authensor/sentinel
4

Audit Layer

Every decision produces a cryptographic receipt. SHA-256 hash chain links receipts together. Tamper with one and the chain breaks.

Hash-chained receiptsNDJSON exportDashboard
5

Governance Layer

Human-in-the-loop approval workflows with multi-party quorum. Webhook notifications, expiration timers, and mobile-friendly approval UI.

Multi-party approvalWebhook notificationsBudget controls
DEPLOYMENT

Run it your way.

Start offline with zero infrastructure. Scale to self-hosted when you need persistence. Upgrade to hosted when you want zero ops.

Zero infrastructure

Offline mode

Policy engine runs in-process. No network calls, no database, no control plane. Perfect for local development, CI, and edge deployments.

$ npm install @authensor/sdk
Full control

Self-hosted

PostgreSQL + control plane on your infrastructure. Docker Compose, Kubernetes, or bare metal. All data stays in your network.

$ docker compose up -d
Zero ops

Hosted

Managed control plane with SLA. Same engine, same API, same receipts. Free tier included for small teams.

$ authensor.com/signup
INTEGRATIONS

Works with every framework.

Drop-in adapters for the major agent frameworks. Or use the SDK directly for custom agents.

LangChain / LangGraph@authensor/langchain
OpenAI Agents SDK@authensor/openai
CrewAI@authensor/crewai
Claude Code (hooks)create-authensor --template claude-hooks
MCP servers (any)@authensor/mcp-server --gateway
Custom agents@authensor/sdk guard()

Ready to get started?

Free, open source, and self-hostable. Works with any agent framework.