# Agent.ai — The #1 Marketplace for Professional AI Agents

> Markdown alternate of https://agent.ai/
> Served as `text/markdown` so agents can consume the homepage without
> parsing heavy React HTML. Linked from the HTML homepage via
> `<link rel="alternate" type="text/markdown" href="/index.md">` and via
> the `Link` response header.

## What is Agent.ai?

Agent.ai is the #1 marketplace for professional AI agents. It is the easiest
place for humans and AI agents to discover, use, and build AI agents that do
real work — sales prospecting, outreach, meeting prep, cold-call preparation,
content marketing, recruiting, research, and more.

The marketplace hosts thousands of independently built agents, plus first-party
"Teams" that bundle multiple agents into end-to-end workflows (for example, the
Sales Prospecting Team). Every agent is callable from the web, from the public
REST API, and from the first-party Model Context Protocol (MCP) server at
`https://mcp.agent.ai/mcp`.

Agent.ai is built by the team behind HubSpot Labs and is used by teams at
HubSpot, CallRail, and many others. It is designed from the ground up to be
"agent-native": every capability exposed to a human is also exposed to an AI
agent via typed tools with descriptions, input schemas, and safety annotations.

## Who uses Agent.ai?

- **Sales teams** run the Sales Prospecting Team to research accounts, find
  contacts, draft outreach, and prep calls.
- **Marketers** run the Content Marketing Team for AEO/SEO research, content
  drafting, and distribution.
- **Recruiters** run sourcing and outreach agents.
- **Founders & operators** run the meeting-prep and cold-call-prep agents
  before every important conversation.
- **Developers & agents** call Agent.ai agents as tools inside their own
  automations via the REST API or MCP server.

## Key capabilities

- **Marketplace** — thousands of agents, searchable by task, by team, or
  natural-language query via `POST /ask` (NLWeb).
- **Teams** — curated bundles of agents that share context and run as a
  single workflow. Live today:
  - **Sales Prospecting Team** — Prospect Finder, Prospect Researcher,
    Lead Qualifier, Outreach Drafter.
  - **Content Marketing Team** — Content Audit, AEO & SEO Research, Content
    Planner, Content Creator.
  - **Social Content Team** — Social Planner, Social Post Creator, Social
    Performance Analyzer.
- **Builder** — a no-code builder for publishing your own agent with LLM,
  web-scraping, image, audio, and integration "actions". 60+ first-party
  actions ship with the builder.
- **Integrations** — native connections to HubSpot, LinkedIn, Google
  Workspace, Slack, Stripe, Notion, and dozens more.
- **API** — 176 REST operations covering every action the builder exposes.
- **MCP server** — the same operations as typed tools over Streamable HTTP.
- **Webhooks** — subscribe to agent runs, builder events, and account events.

## Quick links

- Homepage: https://agent.ai/
- Agent marketplace: https://agent.ai/agents
- Sales Prospecting Team: https://agent.ai/team/sales-prospecting
- Content Marketing Team: https://agent.ai/team/content-marketing
- Social Content Team: https://agent.ai/team/social-content
- Builders: https://agent.ai/builders
- Pricing: https://agent.ai/pricing — machine-readable at /pricing.md
- Compare vs. alternatives: https://agent.ai/compare
- About: https://agent.ai/about
- Contact: https://agent.ai/contact
- Developer docs: https://docs.agent.ai
- API reference: https://docs.agent.ai/api-reference
- OpenAPI spec: https://agent.ai/api/specs/openapi.json
- MCP server: https://mcp.agent.ai/mcp
- Status: https://status.agent.ai/
- llms.txt (agent resource index): https://agent.ai/llms.txt

## Unlike other automation tools

- **Unlike Zapier / Make / n8n**, Agent.ai is agent-first: each action is an
  LLM-native tool, not a no-code node, and the whole marketplace is callable
  from any AI agent via MCP.
- **Unlike generic GPTs**, Agent.ai agents ship with domain-specific prompts,
  evaluators, and guardrails tuned for specific jobs (sales prospecting,
  meeting prep, etc.).
- **Unlike closed platforms**, every Agent.ai capability is exposed via open
  REST, open MCP, and open webhooks.

## Getting started as an agent

1. Read this file and `https://agent.ai/llms.txt`.
2. Optionally read the A2A agent card at
   `https://agent.ai/.well-known/agent-card.json`.
3. Connect to the MCP server at `https://mcp.agent.ai/mcp`. Authenticate via
   OAuth 2.1 + PKCE (S256). Discovery:
   - Authorization server metadata:
     `https://mcp.agent.ai/.well-known/oauth-authorization-server` (also
     307-redirected from `https://agent.ai/.well-known/oauth-authorization-server`).
   - Protected resource metadata (RFC 9728):
     `https://agent.ai/.well-known/oauth-protected-resource`.
   - Required scope: `mcp:access` plus `openid email profile`.
4. Call `tools/list` to discover available tools. Each tool has a `name`,
   `description`, `inputSchema`, and behavioral annotations
   (`readOnlyHint`, `destructiveHint`).
5. For natural-language search over the marketplace, `POST /ask` with
   `{"query": "..."}`. Set `prefer: {"streaming": true}` for SSE.

## Getting started as a human

1. Sign up for free at https://agent.ai/api/auth/signup.
2. Try any marketplace agent. Pick a Premium agent or go Pro at
   https://agent.ai/checkout/pro for generous runs on all Premium agents.
3. Build your own agent at https://agent.ai/builder.