# uber/ADR

ADR secures enterprise AI agents through observability, security benchmarking, and threat detection. Deployed at Uber.

Repository: https://github.com/uber/ADR
Canonical: https://ross.abutalabs.com/products/adr
Homepage: https://arxiv.org/abs/2605.17380
Language: Python
License: Apache-2.0
License Family: permissive
Topics: agent-security, ai-agents, ai-security, benchmark, llm-security, mcp, model-context-protocol, prompt-injection, threat-detection, claude, claude-code, codex, cursor
Last push: 2026-08-26T18:17:48+00:00

## Health v2 (maintenance only)
Score: 68/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 63, longevity 9
- inputs: {"age_days": 136, "days_push": 7, "days_rel": 33, "gap_med": null, "n_releases_24m": 1}
- flags: young
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1501, forks 136 (observed 2026-08-28T04:04:54.311488+00:00)

## What it is
ADR (Agentic AI Detection and Response) is an open-source enterprise security framework for AI agents, providing endpoint discovery of AI tools, agent telemetry collection, a security benchmark (ADR-Bench), and two-tier threat detection for MCP-based agents. It is deployed in production at Uber and accompanies an MLSys 2026 paper.

## Use cases
- detect prompt injection and unsafe actions in AI coding agents
- inventory AI apps and MCP servers on employee endpoints
- collect telemetry from Claude Code, Cursor, and Codex
- benchmark agent security against attack techniques
- detect credential exposures by AI agents
- secure enterprise AI support agents

## When to choose
- you need observability and threat detection for employee-facing AI agents
- you want to benchmark MCP server and agent security
- you run an enterprise security team adopting AI coding tools

## When to avoid
- you need the prevention component, which is not open-sourced
- you need a lightweight consumer tool rather than enterprise deployment
- you only need generic EDR without agent-aware telemetry

## Facets
- artifact type: framework
- maturity: active
- function: security, monitoring, benchmarking, mcp, agent-framework, logging
- domain: security, large-language-models, developer-tools, monitoring
- platform: windows, python, cross-platform
- tags: agent-security, prompt-injection, llm-security, threat-detection, mcp-security, ai-agent-observability, security-benchmark, edr, ai-agents, macos, linux

## Member repositories
- uber/ADR (main) score 68

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:54.311488+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:32:57.904209+00:00, confidence not recorded.
  - readme: https://github.com/uber/ADR (fetched 2026-08-28T04:04:54.311488+00:00, sha f29b56473563)
  - homepage: https://arxiv.org/abs/2605.17380 (fetched 2026-08-29T11:38:00.532185+00:00, sha fe725c305c22)
  - site_page: https://info.arxiv.org/about/donate.html (fetched 2026-08-29T11:38:00.541182+00:00, sha cca9c3a11c56)
  - site_page: https://info.arxiv.org/about/ourmembers.html (fetched 2026-08-29T11:38:00.544418+00:00, sha 47cbc55ff1de)
  - site_page: https://info.arxiv.org/about (fetched 2026-08-29T11:38:00.546246+00:00, sha a1f16f915a9a)
  - site_page: https://info.arxiv.org/labs/index.html (fetched 2026-08-29T11:38:00.542905+00:00, sha b14a8d05a0ec)
- Data as of 2026-08-30T08:39:29.467469+00:00.