Exa

What do you like best about the product?

The quality of the search results is easily the standout feature. When you’re feeding data into LLMs, the output is only as good as the retrieval step, and Exa’s neural search consistently surfaces high-signal content that traditional APIs tend to miss. On top of that, performance feels incredibly snappy, and the API documentation made onboarding straightforward. Overall, it’s a major upgrade for any AI-driven search workflow.

What do you dislike about the product?

When you’re building autonomous agents that run in continuous loops, rate limits and API costs can quickly become a bottleneck as you scale. It’s fantastic for prototyping, but once you deploy agents that need constant, high-volume context retrieval, you have to pay close attention to optimization to keep usage costs under control.

What problems is the product solving and how is that benefiting you?

We struggled to feed high-quality, highly relevant context into our autonomous coding agents. When we were building reinforcement learning environments or SWE-agents, traditional search APIs produced too much noise and often returned poorly formatted HTML. Exa addresses this by offering semantic, neural search that delivers clean, structured data directly to our LLMs. The result is noticeably better agent accuracy, along with hours saved because we no longer have to build and maintain custom web scrapers.