Agentlas Network: borrowing agents from the Hub
Owning your own agents is half the picture. The other half is being able to borrow someone else's expertise for a job, without losing the ownership guarantees above.
What borrowing means here
The Hub is a directory of published agents other people built and shared. Borrowing one does not mean Agentlas runs that agent on a server on your behalf — it means your local runtime downloads that agent's instructions and skill definitions, attaches them to your own project context, and runs them on your own connected AI account. Execution stays local; only the agent's packaged instructions travel.
Search order
- Your own Cloud-saved agents first.
- Your bookmarked agents next.
- The public Hub last, if nothing local fits.
What a borrowed agent can and can't see
A borrowed agent runs with the permissions declared in its package manifest — it does not get blanket access to your files or memory. The publisher's own source, private notes, and secrets are never included in what you borrow; only the packaged instructions and declared tool requirements are.
The mechanism that matches a request to the right agent across multiple runtime environments — routing cards plus non-identifying routing receipts, not raw prompts — is covered by a filed, pending patent application.
Credits and the publisher economy
Most agent marketplaces either don't pay individual creators, or require enterprise-scale partnerships to list. Agentlas Network pays any publisher a share whenever their agent is borrowed and run.