23 open-source AI Cowork vendors
Open-source software for AI Cowork that you can self-host, with auditable source code — regardless of the developer's headquarters.
- vendors scored
- 23
- European
- 1
- last checked
- 11 October 2026
Vendors, scored
Score per vendor: 0 = high exposure, 100 = sovereign
Sort
- Sovereignty score 92 out of 100Nextcloud AssistantGermany · Self-hosted / open sourceOpen-source AI assistant built into the Nextcloud collaboration platform (Stuttgart, Germany), enabling on-premises/self-hosted content generation, summarization, translation and document analysis with open-source LLM backends, avoiding third-party data exposure. Actively offered as part of Nextcloud Hub in 2026.
- Sovereignty score 79 out of 100Pydantic AIUnited Kingdom · Self-hosted / open sourceOpen-source Python agent framework from the Pydantic team (Pydantic Services Inc.) with typed outputs and durable execution; Logfire has an EU region.
- Sovereignty score 77 out of 100EigentUnited Kingdom · Self-hosted / open sourceOpen-source (Apache-2.0) cowork desktop app for a team of AI agents, from Eigent UK Ltd, the UK company behind CAMEL-AI. Runs locally or via the vendor cloud.
- Sovereignty score 71 out of 100LlamaIndexUnited States · Self-hosted / open sourceOpen-source Python framework from LlamaIndex Inc. (US) for agentic applications over documents and your own data; LlamaParse cloud platform with an EU region.
- Sovereignty score 71 out of 100Microsoft Agent FrameworkUnited States · Self-hosted / open sourceOpen-source framework from Microsoft (US) for AI agents and multi-agent workflows in .NET and Python; successor to Semantic Kernel and AutoGen.
- Sovereignty score 71 out of 100OpenAI Agents SDKUnited States · Self-hosted / open sourceLightweight open-source SDK from OpenAI (US) for multi-agent workflows with handoffs, guardrails and tracing; usable with other model providers too.
- Sovereignty score 71 out of 100RivetUnited States · Self-hosted / open sourceUS open-source (MIT) visual IDE and TypeScript library for AI agents, developed by Ironclad.
- Sovereignty score 71 out of 100smolagents (Hugging Face)United States · Self-hosted / open sourceCompact open-source Python library from Hugging Face (US, with an EU entity in Paris) for agents that write their actions as code.
- Sovereignty score 69 out of 100AgnoUnited States · Self-hosted / open sourceOpen-source Python framework (Apache-2.0) and runtime for multi-agent systems that runs agents in your own cloud; formerly Phidata, based in New York, US.
- Sovereignty score 69 out of 100Browser UseUnited States · Self-hosted / open sourceOpen-source library (Python/TypeScript) from Browser Use, Inc. (US) that lets AI agents operate a web browser; with an optional cloud service.
- Sovereignty score 69 out of 100Google Agent Development Kit (ADK)United States · Self-hosted / open sourceOpen-source, code-first framework from Google (US) for building, evaluating and deploying AI agents in Python, TypeScript, Go, Java and Kotlin.
- Sovereignty score 69 out of 100LangflowUnited States · Self-hosted / open sourceOpen-source low-code builder for AI agents and RAG applications with visual flows; part of IBM (via DataStax), US
- Sovereignty score 69 out of 100LettaUnited States · Self-hosted / open sourceOpen-source platform from Letta (US, San Francisco) for AI agents with persistent memory; Apache 2.0 licence, runnable on own infrastructure or as a hosted service.
- Sovereignty score 69 out of 100MastraUnited States · Self-hosted / open sourceOpen-source TypeScript framework from Mastra (US, San Francisco) for AI agents and workflows, with memory and observability; Apache 2.0 core, enterprise modules licensed separately.
- Sovereignty score 69 out of 100OpenClawUnited States · Self-hosted / open sourceOpen-source (MIT) personal AI assistant that runs on your own computer and does tasks through chat apps; stewarded by the US-based OpenClaw Foundation.
- Sovereignty score 69 out of 100PraisonAIUnited States · Self-hosted / open sourceOpen-source AI agent framework with multi-agents, memory and RAG, actively maintained by an individual developer (Mervin Praison) with no company behind it.
- Sovereignty score 57 out of 100Rowboat (Rowboat Labs)India · Non-EU vendorOpen-source personal AI assistant (Apache-2.0) from Y Combinator company Rowboat Labs (Bengaluru, India), local-first with Markdown storage.
- Sovereignty score 49 out of 100AutoGen (Microsoft)United States · Non-EU vendorOpen-source multi-agent framework from Microsoft Research where agents collaborate through conversation; in maintenance mode since October 2025. Controlled by Microsoft, not a community-governed project.
- Sovereignty score 49 out of 100CrewAIUnited States · Non-EU vendorOpen-source Python framework for orchestrating autonomous AI agents that collaborate to solve complex tasks; backed by a real company (CrewAI Inc.), not a controlling-company-free project.
- Sovereignty score 49 out of 100DifyUnited States · Non-EU vendorOpen-source LLM application building platform (LangGenius Inc.) with RAG, agent capabilities and observability, deployable on cloud, VPC, or self-hosted.
- Sovereignty score 49 out of 100LangGraph (LangChain)United States · Non-EU vendorLangChain's open-source graph-based framework for building stateful AI agents with explicit state management, checkpointing and human-in-the-loop capabilities; controlled by LangChain Inc., not a community-governed project.
- Sovereignty score 49 out of 100SuperAGIUnited States · Non-EU vendorOpen-source autonomous AI agent framework for developers; core development has slowed since 2024 as the company pivots to a commercial SaaS offering.
- Sovereignty score 49 out of 100VoltAgentUnited States · Non-EU vendorOpen-source, TypeScript-first AI agent framework with a workflow engine, multi-agent orchestration, and a coupled observability console (VoltOps).
Assess your own stack in the assessment
Assess your stack →