cuga-project/cuga-agent

CUGA is an open-source generalist agent harness for the enterprise, supporting complex task execution on web and APIs, OpenAPI/MCP integrations, composable architecture, reasoning modes, and policy-aware features.

What it solves

CUGA is an enterprise-grade agent harness designed to eliminate the need to build domain-specific AI agents from scratch. It provides a pre-built orchestration layer, planning logic, and safety policies, allowing developers to focus on configuring domain-specific tools, workflows, and policies rather than rebuilding the core agentic infrastructure.

How it works

CUGA functions as a configurable generalist agent that combines planner-executor and code-act patterns with structured planning. It integrates with various tool ecosystems including OpenAPI specs, MCP servers, and LangChain. For complex tasks, it supports a hybrid mode that combines API calls with browser interactions via Playwright. It also features a policy system for governance (e.g., Intent Guard, Tool Approval) and a knowledge engine for RAG using Docling to ingest various document formats.

Who it’s for

It is designed for enterprise developers and organizations that need to deploy reliable, domain-specific agents quickly while maintaining strict control over behavior, safety, and deployment (e.g., via Kubernetes).

Highlights

  • High Benchmark Performance: Ranked #1 on AppWorld and WebArena benchmarks.
  • Flexible Tooling: Supports MCP, OpenAPI, and LangChain tool integration.
  • Enterprise Governance: Includes a policy system with human-in-the-loop approval gates.
  • Hybrid Execution: Can seamlessly switch between API-based and browser-based task execution.
  • Knowledge Integration: Built-in RAG with agent-level and session-level document scoping.
  • Modular Deployment: Supports self-hosting on Kubernetes via Helm charts and integrates with Langflow for low-code visual workflows.

Related

  • Dispatch
    CUGA on Hugging Face: Democratizing Configurable AI Agents

    IBM Research has released CUGA (Configurable Generalist Agent), an open-source agent framework that achieves state-of-the-art performance on AppWorld and WebArena benchmarks for complex API and web tasks.

  • Project
    trycua/cua

    An open-source infrastructure suite that provides isolated cloud and local desktops, desktop automation tools, and benchmarks for AI agents to operate computers.

  • Dispatch
    OpenAI Computer-Using Agent (CUA) and Operator Research Preview

    OpenAI has introduced Computer-Using Agent (CUA), a model capable of interacting with GUIs via a universal interface of screen, mouse, and keyboard to perform digital tasks, now available in research preview via Operator.

  • Project
    java-up-up/nexus-agent

    An enterprise-grade AI agent platform that implements a complete loop from document governance and hybrid RAG retrieval to ReAct-based agentic reasoning and MCP tool integration.

  • Project
    dograh-hq/dograh

    Dograh AI is an open‑source, self‑hostable platform for building and testing voice conversational agents. It provides a visual workflow editor, telephony integrations (Twilio, Vonage, etc.), and a modular LLM/STT/TTS stack that you can replace with any provider. One‑command Docker setup gets the whole system running locally, and Python/Node SDKs let you create agents programmatically. The project is BSD‑2 licensed and backed by a Slack community.