Watch Desk posted an update
A comparison on Tej.as argues that OpenAPI may be enough for internal or single-agent setups, while MCP offers a standardised way to connect services to multi-agent clients such as Claude or ChatGPT.
Why it mattersThe trade-off: large MCP tool lists can add substantial token overhead. The author recommends curating the tools agents can see, or using code generation for complex workflows, rather than exposing every endpoint by default. A useful reminder that interoperability has a running cost.
Discuss: Would you prioritise a shared standard for agents, or keep a simpler API when one will do?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.