A phone book for AI agents — 256 services they can actually use

What it does

This is a curated catalog of agent-native services — email, remote browsers, memory, sandboxes, payments, MCP tools, telephony, even social networks where agents are first-class users. It spans 16 categories and 256 entries, each with a tagline, primitives, MCP support, and an onboarding pointer. Crucially, it’s not a list of frameworks for building agents; it only admits infrastructure an agent consumes directly or purpose-built surfaces for running agent sessions.

The interesting bit

The list is machine-readable by design: there’s a skill.md entry point so an agent can browse the catalog itself, plus a Skills Hub with workflows like find-agent-service and install-agent-service that rank onboarding paths (URL first, then plugins, MCP, CLI, SDK). The catalog wants to be an installer/router, not just a directory. There’s also a published selection criteria doc and an explicit “excluded / boundary cases” section — rare discipline for an awesome-list.

Key highlights

• 256 services across 16 categories, from agent email to LLM gateways to agent social networks
• Per-service files with primitives, MCP support, and onboarding guides
• Skills Hub lets agents operate the catalog directly (find, install, vet, contribute)
• A web version with structured metadata and machine-readable discovery endpoints
• Contribution workflow is issue-first, with an evaluate-agent-native skill for vetting candidates

Caveats

• The “How to Use” column mixes onboarding styles — some entries are just “read the docs and follow instructions,” which is less agent-native than the framing suggests
• As with any awesome-list, inclusion is curation, not endorsement; quality across 256 entries will vary

Verdict

If you’re wiring an agent into real-world infrastructure and don’t want to rediscover the landscape service by service, this is a genuinely useful map. If you wanted agent frameworks, it will politely point you elsewhere — that’s the point.