AI Agents
koog
Koog is a JVM (Java and Kotlin) framework for building predictable, fault-tolerant and enterprise-ready AI agents across all platforms – from backend service…
★ 4.6k starsApache-2.0
View guide →Uber's ADR: discover, observe, benchmark, and detect threats across enterprise AI agents, deployed at Uber.
ADR is an open-source ai agents project by uber (uber/ADR on GitHub). It has 1,950 GitHub stars and 202 forks, is released under the Apache-2.0 license, and was last updated on GitHub on 8 October 2026.
Decide which component you need first: endpoint discovery, telemetry collection, or the detection benchmark.
Clone the repository and move into the component you are evaluating, such as Detection.
Sync the Python environment with uv and configure the model credentials the detector needs, or choose the keyless smoke-test detector.
Follow the reproducibility workflow to inflate the benchmark, run detectors, and review results before adapting components to your fleet.
Inventories installed AI applications, CLI agents, IDE extensions, local model runtimes, and MCP servers on an endpoint, flagging unknown surfaces for review.
Collects and normalises agent intent, tool use, and execution traces from many coding tools across macOS, Linux, and Windows into a unified schema.
A security benchmark with hundreds of tasks and many MCP servers covering a range of agent attack techniques, with baselines included.
Combines high-recall triage with deeper agentic reasoning on suspicious sessions to detect risky agent behaviour efficiently.
Benchmark fixtures are synthetic (fake credentials and emulated environments) intended for defensive security research.
A step-by-step workflow documents how to reproduce benchmark detection results and the paper's figures.
The quickstart in the README focuses on ADR Detection.
git clone https://github.com/uber/ADR.git
cd ADR/DetectionThe default adr detector uses an Anthropic key; a keyless smoke-test detector is also documented.
uv sync
export ANTHROPIC_API_KEY=<your key>Follow the reproducibility guide to inflate the packed benchmark and run detectors. # see README docs/REPRODUCIBILITY.md
# see docs/REPRODUCIBILITY.md for the full evaluation workflowIt is developed by Uber, where the README says it is deployed in production, and published under the uber GitHub organisation.
Yes, the released components are Apache-2.0 licensed. Running the default detector uses your own Anthropic API key; a keyless smoke-test detector option is documented.
For the default detector, yes, an Anthropic API key is configured in the quickstart. Organisations can adapt components to their own model setups.
Endpoint security tools watch processes and files; ADR is purpose-built for agent behaviour: it inventories AI tools, observes agent traces, and benchmarks agent-specific attacks such as prompt injection through MCP servers.
Clone the repo, cd into Detection, run uv sync, set ANTHROPIC_API_KEY, and follow docs/REPRODUCIBILITY.md.
Koog is a JVM (Java and Kotlin) framework for building predictable, fault-tolerant and enterprise-ready AI agents across all platforms – from backend service…
A rigorous benchmark for LLMs acting as agents.
A control plane with a durable state kernel that keeps long-horizon agents moving across sessions.
OSWorld — benchmark multimodal agents on open-ended tasks in real computer environments.
Open-source foundation of ToolJet AI - the enterprise app generation platform for internal tools, dashboards, business applications, workflows and AI agents.…
An open benchmark for browser AI agents on everyday tasks.
The most-starred other AI Agents projects in the LibreAI catalogue — the realistic shortlist if ADR is not the right fit.
A personal AI that really does things — on any OS, with your data staying yours.
An agentic skills framework & software development methodology that works.
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
The agent that grows with you — a Python framework for adaptive AI agents.
DeepSeek's plugin-based agent harness where everything is a plugin.
Free and open — grab the source, star the repo, and start creating.