If you came here expecting a proxy review, adjust your expectations. Nimble has moved on.

The product now is expert level web search for agents, and the proxy layer has become plumbing underneath rather than the thing being sold.

That shift is worth understanding, because it is where a lot of this industry is going.

What Is Actually On Offer

Four things, and none of them is an IP address.

The Agent API automates web research tasks end to end. The Search API searches the live web and returns fresh results. The Extract API pulls structured data from a URL. And an MCP server lets an AI system connect to all of it directly.

There are also integrations and AI plugins for slotting the whole thing into an existing stack.

Self Learning Search Agents

The headline feature is agents that learn your specific use case rather than running a fixed routine.

The practical claim is that accuracy improves as the agent sees more of your domain. Instead of you refining prompts and selectors over weeks, the system adapts.

Treat that as a claim to test rather than a fact, as with anything described as self learning. But the direction is right, and it addresses the part of web data work that genuinely consumes people's time: not fetching pages, but keeping extraction correct as the web changes underneath you.

Why the Proxy Layer Disappeared

Follow the logic and the move makes sense.

A proxy solves one problem: the address you appear to come from. Then you still need rendering, challenge handling, parsing, retries and monitoring. That is the bulk of the work, and it never stops.

Selling the outcome rather than the ingredient means the customer stops maintaining all of that. For a team whose actual job is analysis rather than scraper maintenance, that trade is usually worth making. Our scraping guide lays out what that maintenance actually involves, which is the honest way to judge the price.

The MCP Server

Worth calling out separately. An MCP server means an AI agent can use Nimble as a capability rather than through code somebody wrote to wrap an API.

Very few companies in this space have built this yet. If you are working on agent based systems, it changes web access from an integration project into something simply available.

Who It Fits

Nimble suits teams building AI agents that need reliable web research, organisations that want structured data rather than raw pages, anyone whose extraction code keeps breaking, and work where accuracy matters more than unit cost.

Who Might Look Elsewhere

If you want proxies, this no longer sells them in the way you mean. Buy addresses from a provider that still leads with them.

Cost sensitive bulk collection will also find this expensive compared with raw proxies, because you are buying the whole solved problem rather than one component.

Final Verdict

Nimble is a web intelligence platform that used to be a proxy company, and the transition is instructive.

Agent, Search and Extract APIs, an MCP server, integrations across common stacks, and a focus on the part of this work that is genuinely hard.

If your problem is getting an IP address, look elsewhere. If your problem is that your research pipeline keeps breaking and somebody has to fix it every fortnight, this is the category of answer worth pricing.