Uber ADR

Uber ADR is an enterprise security system that secures agentic AI by providing discovery, observability, threat detection, and security benchmarking.

ADR is a security system designed to protect employee-facing and customer-facing AI agents. It functions by discovering unsanctioned AI tools, observing agent activity across endpoints, and benchmarking agent security against specific attack techniques. The system includes an inventory of installed AI applications, IDE extensions, CLI agents, and local model runtimes. It uses a dual-agent architecture for detecting suspicious behavior. Developed at Uber, the platform supports multiple AI coding tools and integrates with various internal automation and support agents.

Uber ADR at a glance

Best for: Software engineer
Runs on: macOS, Linux, Windows, open source
Pricing: Open source under Apache License 2.0
Released: April 2026
Category: AI coding tool
Website: github.com/uber/ADR

Key features

Inventory of AI applications, CLI agents, IDE extensions, and model runtimes
Normalization of agent telemetry across multiple AI coding tools
Security benchmarking using over 300 tasks and 134 MCP servers
Dual-agent architecture for behavior detection
Endpoint discovery of MCP servers
Telemetry capture of tool use and execution traces

Use it for

Discovering unsanctioned AI tools on employee endpoints
Testing agent security under enterprise conditions
Detecting risky agent behavior in production sessions
Collecting normalized telemetry from AI coding assistants

Alternatives to Uber ADR

ARTEX — Runs autonomous multi-agent penetration testing with a dual-graph architecture.
Arthur — Enforces policies and monitors enterprise AI agents on a single control plane.
Arcjet — Enforces security policies directly at the action boundary inside your application code.
Coder — Provides full security, audit logging, and compliance for AI development.
Aegis Kernel — Stops prompt injection and tool-hijacking at the OS syscall boundary.
ADRs4AI — Captures and version-controls the reasoning behind human-AI collaboration.
Adhdev — Orchestrates agents across machines with automated, test-gated git merging.
Composio — Equips AI agents with tools, managed user auth, and sandboxes across 1,500+ apps
OpenAPPA — Tracks data flows algebraically to prevent prompt injection without breaking agents.
Google ADK — It applies software development principles to agent creation with Python.
Eve — Builds durable AI agents with file system based configuration and isolated sandboxes.
Melra — Governs AI agent tool calls with local policies, approval gates, and verified receipts.

All Uber ADR alternatives