Frameworks, workflows, and workspace agents
AI Agents
Libraries, visual builders, and hosted agents. Observability tools that exist to watch those agents are included when builders use them as part of the same stack. They are not sorted above frameworks.
Bench task: Complete a multi-step job that calls one external tool and produces a checkable result. Record reliability, cost, time, and how long setup took from a clean machine.
Scored
Sorted by the published overall score. A tie breaks A to Z. Sponsorship, affiliates, and fast-track fees are not inputs.
No tested scores in this list yet. The first bench has not been published, so nothing here has a number.
Not yet scored
Alphabetical. This is the testing queue, not a ranking.
- AutoGenPending testMicrosoft's open-source multi-agent frameworkOpen source
- CrewAIPending testOpen-source framework for multi-agent crewsOpen source
- DifyPending testOpen-source platform for LLM apps and agentsOpen source
- DustPending testWorkspace agents connected to company toolsPaid
- FlowisePending testOpen-source visual builder for LLM flowsOpen source
- LangGraphPending testGraph-shaped, stateful agent workflowsOpen source
- LangSmithPending testTracing and evals for LLM applicationsFreemium
- MastraPending testTypeScript framework for agents and workflowsOpen source
- n8nPending testWorkflow automation with AI agent nodesSource available
- OpenAI Agents SDKPending testOfficial SDK for tools, handoffs, and tracingOpen source
- ZapierPending testApp automation, including agents that take actionsFreemium