one agent that handels environments(handles, api keys, auth, data, schema) another agent to brainstorm, another agent to code, another agent to double check the work and fix typos.

and please can you stop these agents from apologising and telling us "this is the final solution" when it's not. just cut the human ******** and just code, you can continue to explain, but cut the excuses and apologies. they are annoying and waste tokens.