An independent site about getting AI coding agents configured and working. This page explains how it makes money and what rules it follows, because you should know that before you take advice from anyone.
Some links to AgentRouter on this site are referral links. If you sign up through one, we may earn a referral reward. It costs you nothing extra.
That creates an obvious incentive to tell you to switch providers whether or not switching would help you. Here is how we handle it: every fix page puts the free diagnosis first and the referral option last, after the steps that cost nothing. On the usage-limits page in particular, we say plainly that a second endpoint will not fix a session that wastes tokens — which is advice against clicking our own link.
The reason is not altruism. A page that misdiagnoses your problem to earn a referral produces a signup that churns, and it destroys the only thing that makes a site like this work, which is being right often enough to be worth bookmarking.
We do not publish pricing, promotional numbers, rate limits, or model availability that we have not verified on the provider's own page. Those change without notice, and a stale number presented confidently is worse than no number at all — it sends you to configure something that no longer exists.
Where our diagnostic content carries a factual claim, it is tagged with its source. Anything not yet verified against a primary source renders with a visible 'unverified' badge rather than being quietly presented as fact. If you see a badge, treat that item as a lead to check, not as an answer.
The Setup Doctor is a deterministic rule engine, not a chatbot. Your answers are matched against a fixed rule set, the most specific matching rule wins, and the same answers always produce the same diagnosis. There is no model call in the diagnostic path.
This is a deliberate design choice. A diagnostic whose output varies run to run cannot be tested, and an incorrect fix delivered with fluent confidence is worse than a blunt one. Because the rules are fixed, they can be unit tested — and they are, including a test that a precise sourced rule always outranks a broad fallback.
Your answers are recorded to understand which problems people actually arrive with, so the rule set can be improved. We store a salted, non-reversible visitor hash rather than your IP address, and no account or email is required to use the tool.
Not affiliated with, endorsed by, or sponsored by Anthropic, OpenAI, AgentRouter, or any model vendor. Product names are used for identification only.
Before relying on anything here for a spending decision, verify current pricing and model availability on the provider's own site. We are a starting point for diagnosis, not a source of truth about someone else's commercial terms.