Download app now 
Category
Create AI Agent
Platforms and building blocks for AI agents that act rather than just answer: no-code agent builders, customer service agents, coding agents, sandboxes for running agent code, and web search for agents.
An agent is a language model given tools and permission to use them: it searches, reads company data, writes and runs code, and decides its own next step. The tools here cover different layers of that job. Some, like Dify and Dust, let you assemble an agent without code, the latter connecting it to Notion, Slack and Google Drive. Others are infrastructure the agent runs on.
Read more about this category
Give the agent a sandbox, not your server
Once an agent executes code, the question is where. E2B and Daytona provide isolated environments so generated code runs without touching production. Pair that with logging of every call the agent makes, which is exactly what you want when it does something nobody expected. Exa covers another common need, giving an agent semantic search over web sources.
Customer-facing or internal?
Customer service platforms such as Sierra and Decagon should be judged on their guardrails and on whether the agent can act inside your systems, not merely reply. Test them in every language and dialect your customers actually write in. Internal engineering agents like Factory write, review and deploy code, so decide up front which changes need human sign-off. For the underlying models, see GPT.
