Anthropic Claude Agent SDK Integration

The Anthropic Claude Agent SDK is a developer framework by Anthropic that enables the programmatic creation and deployment of autonomous AI agents capable of multi-turn reasoning, executing code, and interacting with file systems and external tools.

Start for free

Anthropic Claude Agent SDK

Benefits with Ory + Anthropic Claude Agent SDK Integration

Ory integrates with Anthropic's Claude Agent SDK to provide security guardrails for developers and builders to construct applications and agents safely. Ory's harness hook executes an identity and credential check on every session and tool call, enforcing fine-grained authorization to allow or deny actions while logging all activity for an audit trail. Teams can use Observe mode to monitor how developers and builders use tools, or Enforce mode to strictly restrict agents to their assigned permissions.

The fastest way to get started with Ory Agent Security is to sign up for the Ory Network. Installing an Ory Agent Security SDK for your AI agent takes seconds. You simply install the package and initialize it in your project.

npm install @ory/claude-agent-sdk

View on npm

Use Cases

— 03 use cases

Agent IAM

Secures Anthropic Claude Agent SDK sessions, tool execution, and autonomous workflows by enforcing fine-grained authorization, identity checks, and complete audit logging.

Explore Agent IAM

CIAM

Accelerates embedding customer authentication and identity management into consumer-facing applications built with Anthropic Claude Agent SDK using Ory best practices.

Explore CIAM

B2B IAM

Streamlines building enterprise SSO, multi-tenant organization structures, and role-based access control into business software, scaling seamlessly with business growth.

Explore B2B IAM

Core Functionality

— 06 capabilities

Identity
Verifies developers and Anthropic Claude Agent SDK agent identities before session initiation across your ecosystem through a unified control plane.
Authorization
Enforces fine-grained, policy-based access control for every Anthropic Claude Agent SDK tool execution and action.
Shell
Intercepts and secures terminal-level shell commands and code execution within the Anthropic Claude Agent SDK environment.
Audit
Logs complete audit trails for all session activity, tool invocations, and API calls.
Rollout
Supports seamless deployment with Observe mode for monitoring agent behavior and Enforce mode for strict permission enforcement.
Security Control
Provides vendor-agnostic governance across all AI tools, complete with instant token revocation to immediately cut off compromised sessions.