haoruilee/awesome-agent-native-services

A curated list of agent-native services and infrastructure for AI agents: email, browsers, memory, sandboxes, payments, and MCP tools. Includes selection criteria and onboarding guides.

What it solves

This project is a curated index of services and infrastructure built specifically for AI agents, not for humans building agents. It solves the problem of discovery: an agent developer or the agent itself needs to find reliable email, browser, memory, sandbox, payment, or other operational services that are designed to be consumed directly by autonomous agents, with native protocols, APIs, and onboarding flows.

How it works

The repository organizes 258 services into 16 capability categories, each with a table listing the service, a one-line tagline, the primitives it offers (such as per-agent inboxes, remote browser sessions, MCP servers, or sandboxes), whether it provides an MCP interface, and a concrete "How to Use" entry point. The catalog is backed by per-service Markdown files that contain detailed onboarding, selection criteria, and admission rules. It also includes a Skills Hub with SKILL.md-based workflows, so an agent can install the repo once and then invoke skills like find-agent-service, install-agent-service, or evaluate-agent-native to discover, connect, or vet a service directly. A skill.md file at the root provides a first-step discovery path for agents, and the same catalog is mirrored on the web with structured metadata.

Who it's for

This is for AI agents themselves, and for developers who build or operate autonomous agents. The README is written in a way that an agent can read it and follow the onboarding instructions. It explicitly distinguishes agent-native infrastructure (services an agent consumes directly) from generic platforms meant for humans to build agents with, so the catalog stays narrowly scoped. It also serves as a contribution hub where maintainers can propose new services using the issue-first workflow.

Highlights

  • 258 services across 16 categories, including email, browser automation, MCP tools, payments, sandboxes, memory, search, voice, and social layers for agents.
  • Each entry lists concrete primitives and a direct "How to Use" command, so an agent can act on the catalog without extra browsing.
  • Includes a Skills Hub with SKILL.md-based skills (find-agent-service, install-agent-service, evaluate-agent-native, add-to-awesome-list) that turn the directory into an installer/router, not just a static list.
  • Provides a Claude Code plugin marketplace installation path and a manual SKILL.md copy workflow for agents that support SKILL.md.
  • Offers a web mirror at the Agent-Native Index for structured discovery and machine-readable endpoints.
  • Includes explicit selection criteria and admission tracks for vetting new services before they are added to the list.

Related

  • Project
  • Project
  • Project
  • Project