BetaFindAgent is in free public beta — every agent is free to connect and paid agents aren't available yet.
Getting started
How FindAgent works
FindAgent is a cross-LLM marketplace of focused AI agents. Find one, connect it to the AI app you already use, and put it to work — in three steps.
GA4 Detectoran agent
FindAgentgateway
Claudeyour client
Step 1
Find an agent
Browse the marketplace by industry or discipline, or search for what you need. Each agent has a public page with what it does and who built it. Looking for a remote MCP server specifically? Browse the MCP directory.
Step 2
Connect it to your AI app
Every agent works the same way across clients. Open an agent and follow its steps — for a hosted, doer, or code agent, connect it by adding its MCP server URL; for a local recipe, install it with the FindAgent CLI — to wire it into:
Once connected, the agent runs where you already work — no new tool to learn. Start a fresh conversation and put it to use.
Two ways to browse
The marketplace is organized on two axes, and an agent can be tagged on both:
Industry— the vertical an agent serves: Healthcare, Finance, Retail, Real Estate, and more.
Discipline— the kind of work it does: Customer Service, HR, Cybersecurity, Legal, Software Development, and more.
Start from the Browse page by Industry or by Discipline, then narrow with the dual-lens header: pick an Industry lens and a Discipline lens and you see their intersection — “Sales agents for Real Estate.” Either lens can stay “Any.” When the active discipline is Software Development, two more refinements appear — Tech Domain (Web, Mobile, Backend, …) and Language/Framework (TypeScript, Python, React, …).
What is an agent here?
Agents on FindAgent come in a few shapes:
Prompt recipe.A packaged system prompt and example flow — instructions your AI app follows.
Tool-using (doer) agents.Agents that call real tools — fetch your analytics, build a spreadsheet, manage a store.
Skills bundles.A Claude, Cursor, Windsurf, or Continue skills-or-rules repo turned into one agent — each skill becomes a prompt your client can call.
Code agents.A creator's real GitHub code, run 1:1 in an isolated sandbox after review. FindAgent reads the repo as-is and generates the agent from it — the creator never writes a config file.
MCP servers.A remote MCP server someone already hosts, listed so you can discover and connect it.
Departments.Two to sixteen agents (up to eight for peer-to-peer) composed into one team that runs as a single MCP server.
Building one yourself? Publish in the submit wizard — or skip the forms and submit straight from your AI client over MCP, where your own assistant reads your repo and writes the listing.