Useful infrastructure, clearly described.
What belongs here
Tools that help teams build and operate AI agents: sandboxes, browser infrastructure, search and web data, memory, application integrations, agent authentication, observability, and evaluations.
We include hosted services and maintained software that runs in your infrastructure. A listing needs a clear infrastructure purpose, public documentation, and an identifiable provider or project. We look for practical setup guidance and evidence of ongoing maintenance.
What we check
Descriptions, category assignments, deployment options, and runtime license labels use official product documentation or repositories. Details & sources on each listing links to the supporting material and shows the review date. A listing is not a certification or a hands-on performance test. Where a project explicitly describes the listed software as beta, we display a Beta label. An absent label does not certify stability.
Hosted, self-hosted, and open source
Hosted means a documented managed service is available. Self-hosted means a documented option runs in customer infrastructure. This may require a commercial plan. Partial deployments, such as a self-hosted data plane with a managed UI, are labeled accordingly.
Open source describes the reviewed license of the listed runtime or framework. Source available indicates a public codebase with additional license restrictions. An open-source SDK does not make a hosted platform open source. If a deployment mode or runtime license is not established, we leave it unlisted rather than treating it as unavailable.
Ordering and alternatives
Tools are listed alphabetically by default. Hosted and self-hosted filters show views of the same catalog, so a tool offering both modes appears once in All tools. Open source is a separate filter.
Alternatives pages list other tools sharing the relevant capability category. They help you discover products to investigate. They do not rank products, compare prices, or claim that features and integrations are equivalent.
Keeping the directory useful
We currently focus on infrastructure rather than finished agents, agencies, generic agent builders, or tutorials. Model gateways and inference are outside the initial scope. If something is missing or outdated, send us a suggestion.