cuga
CUGA is an open-source generalist agent for the enterprise, supporting complex task execution on web and APIs, OpenAPI/MCP integrations, composable architecture, reasoning modes, and policy-aware features.
Decision gist · record as of 2026-08-14
Yes, if you need an enterprise agent harness with built-in tool orchestration, policy controls, and multi-agent supervision. The active maintenance, permissive license, low install friction, and zero known vulnerabilities support adoption. Start with a quick evaluation on your domain tasks—the benchmarks (AppWorld, WebArena) are strong but domain-specific, so verify reasoning modes and cost tradeoffs fit your workload. Experimental features (save-and-reuse, skills sandbox) are worth monitoring but not blockers for production use of core capabilities.AI-flagged interpretation of the facts on this page — verify before relying
Before you install
- Requires Python 3.10+.
- Hybrid mode (web + API tasks) requires Playwright and Chromium: `playwright install chromium`.
- MCP server integration and knowledge features require additional configuration via settings.toml or environment setup.
License · maintenance · safety
Apache-2.0 (permissive) — Apache-2.0 permissive license allows commercial use, modification, and distribution with minimal restrictions, suitable for enterprise deployment and integration.
last release 2026-08-03 (11 days) · last repo commit 2026-08-14 · 867 stars
0 known vulnerabilities (OSV.dev, 2026-08-14) · 168,047 downloads/mo, #10,457 on PyPI
Alternatives
Verify before relying
pip install cuga
from cuga import CugaAgent
agent = CugaAgent(tools=[...])
result = agent.run("your task here")- Actual performance on your specific domain tasks (benchmarks cited are AppWorld and WebArena, not general-purpose workloads)
- Cost and latency tradeoffs across the three reasoning modes (fast/balanced/accurate) in production
- Stability and maturity of experimental features (save-and-reuse, agent skills sandbox modes)
- Kubernetes deployment complexity and operational overhead for self-hosting
What it is and what it does
CUGA is a configurable agent harness designed to reduce the engineering burden of building domain-specific enterprise agents. Rather than orchestrating tools, planning logic, and safety policies from scratch, you configure CUGA with your APIs (via OpenAPI specs or MCP servers), tune reasoning modes (fast/balanced/accurate), and layer policies (Intent Guard, Playbook, Tool Approval, Tool Guide, Output Formatter) to govern behavior. It combines web automation (Playwright) and API calls in a single task, supports multi-agent supervision, and includes a knowledge engine for RAG over ingested documents.
The framework ships with a web UI for drafting and publishing agent configurations, Kubernetes deployment templates for self-hosting, and integrations with LangChain, LangFlow, and external agents. It targets enterprises that need a generalist starting point rather than a blank-slate agent framework, with built-in support for human-in-the-loop approval gates and policy enforcement. Maintenance is active, and the package carries 48 runtime dependencies spanning LLM providers (OpenAI, IBM, Groq, LiteLLM), async HTTP clients, browser automation, and observability tools.
Use it for
- Automate multi-step workflows combining web form filling and backend API calls without custom orchestration code
- Deploy a domain-specific agent with safety policies (e.g., tool approval gates, output formatting rules) via configuration rather than reimplementation
- Build a multi-agent supervisor that routes tasks to specialized agents or external A2A services and aggregates results
- Ingest and reason over enterprise documents (PDFs, Office files, HTML) scoped to agent-level or per-session knowledge contexts
- Prototype and iterate on agent behavior in a web UI, then publish versioned configurations to production chat or Kubernetes clusters
- Integrate custom tools and MCP servers into a reasoning agent without rewriting core orchestration or planning logic
Worth the install?
AI-flagged interpretation of the facts on this page. Verify before relying on it.
Yes, if you need an enterprise agent harness with built-in tool orchestration, policy controls, and multi-agent supervision.
The active maintenance, permissive license, low install friction, and zero known vulnerabilities support adoption. Start with a quick evaluation on your domain tasks—the benchmarks (AppWorld, WebArena) are strong but domain-specific, so verify reasoning modes and cost tradeoffs fit your workload. Experimental features (save-and-reuse, skills sandbox) are worth monitoring but not blockers for production use of core capabilities.
Install
cuga on PyPI
Before you install
Low friction install with a pure Python wheel. Active maintenance (last commit 2026-08-14, 867 repository stars). Requires Python 3.10 or later and manages 48 runtime dependencies including LangChain, FastAPI, Playwright, and Docker—a substantial but well-established ecosystem.
Requires Python 3.10+. Hybrid mode (web + API tasks) requires Playwright and Chromium: `playwright install chromium`. MCP server integration and knowledge features require additional configuration via settings.toml or environment setup.
License in practice
Apache-2.0 permissive license allows commercial use, modification, and distribution with minimal restrictions, suitable for enterprise deployment and integration.
Quickstart
pip install cuga
from cuga import CugaAgent
agent = CugaAgent(tools=[...])
result = agent.run("your task here")Verify before relying
- Actual performance on your specific domain tasks (benchmarks cited are AppWorld and WebArena, not general-purpose workloads)
- Cost and latency tradeoffs across the three reasoning modes (fast/balanced/accurate) in production
- Stability and maturity of experimental features (save-and-reuse, agent skills sandbox modes)
- Kubernetes deployment complexity and operational overhead for self-hosting
Package facts
| License | Apache-2.0 permissive |
| Python support | Supports the current Python release <3.14,>=3.10 |
| Install friction | Low. Pure-Python wheel |
| Runtime dependencies | 48 packageslangchainlangchain-ibmlangchain-openailangchain-coreaiohttphttpxdockerlanggraphmarkdownifyuvicornfastapilangfusedynaconfplaywrightbrowsergym-corepython-dotenvlogurumcppsutiltypertyper-slimfastmcplangchain-mcp-adapterspyyamllangchain-groqlangchain-litellmlitellmtavily-pythonfastembeda2a-sdk |
| Maintenance | Actively maintained 11 days since the last release |
| Last repo commit | |
| First released | |
| Downloads | 168,047 / month, #10,457 on PyPI 30-day window, as of 2026-08-14 |
| Known vulnerabilities | None known OSV.dev, checked 2026-08-14 |
Evidence: cuga-0.3.1-py3-none-any.whl
Tags
Let your AI agent find packages like this
Example. Real query, live index.
You found this page by searching. An agent finds it by wishing: SkillFed indexes 14,416 PyPI packages by what they can do, searchable in plain language.
wish › “configurable reasoning agent”
- cugaCUGA is an enterprise-grade generalist agent framework that…
- agentscopeAgentScope is a production-ready framework for building multi-agent…
- hindsight-api-slimHindsight-api-slim provides a persistent memory system for AI agents…
Give your agent the search over MCP, or paste the wish link into any chat.
More Artificial Intelligence packages
LiteLLM provides a unified Python interface to call 100+ LLM providers (OpenAI, Anthropic, Gemini, Bedrock, Azure, and others) using OpenAI-compatible API format, available as both a Python SDK and a self-hosted AI Gateway proxy server.
Install it if you need to work with multiple LLM providers or want to centralize LLM routing in your organization.
Client library and CLI tool for downloading, uploading, and managing models, datasets, and repositories on the Hugging Face Hub platform.
Install it if you work with Hugging Face Hub models or datasets.
LangChain provides a framework for building agents and LLM-powered applications by composing language models, tools, and memory through a unified API that abstracts over multiple model providers.
hf-xet provides chunk-based deduplication and efficient file transfer for the Hugging Face Hub, enabling faster uploads and downloads of large files with local disk caching.
Tokenizers converts raw text into token sequences for NLP models, with support for training custom vocabularies and using pre-built tokenizers (BPE, WordPiece) optimized for speed via Rust.
Transformers provides a unified framework for loading, fine-tuning, and running state-of-the-art pretrained models across text, vision, audio, video, and multimodal tasks using PyTorch, JAX, or TensorFlow.
Install it if you need to run or train any transformer-based model for NLP, vision, audio, or multimodal tasks.
See also agentscope · google-agents-cli · gpt-researcher · fast-agent-mcp · praisonaiagents · deepagents · nvidia-nat · agent-framework-core · echo-agent · metagit-cli