Agree on flexibility, you can just do a lot more with AI agent + curl + SKILL.md.
The agent comes with a good amount of API knowledge in training data already. A curated set of MCP tools throws most of that away. With curl, it can hit any endpoint and write scripts. SKILL.md covers what it gets wrong (API version, rate limits, etc.) and points at the vendor's own API docs for anything else.
The annoying part isn't tool calling, it's auth. You don't want the agent managing OAuth tokens for 10+ apps. A passthrough proxy that holds the credentials, injects them per request, and scopes permissions solves that. At Maton, we have our own MCP servers as well, but most of our customers use us for managed auth.
byungkyu · · focus · HN ↗
The agent comes with a good amount of API knowledge in training data already. A curated set of MCP tools throws most of that away. With curl, it can hit any endpoint and write scripts. SKILL.md covers what it gets wrong (API version, rate limits, etc.) and points at the vendor's own API docs for anything else.
The annoying part isn't tool calling, it's auth. You don't want the agent managing OAuth tokens for 10+ apps. A passthrough proxy that holds the credentials, injects them per request, and scopes permissions solves that. At Maton, we have our own MCP servers as well, but most of our customers use us for managed auth.
danielpopame · · focus · HN ↗
[dead]