Provider breakage is the silent killer of AI agents

Point an agent at a single LLM provider and you have inherited that provider’s worst day. Free and free-tier endpoints are the most volatile of all: quotas reset unpredictably, models get retired with little notice, and a few even misreport what they can do.

The three ways providers break

• Rate limits. Free tiers cap requests per minute and per day. Hit the cap and every call 429s until the window resets.
• Retirement. A model id you depend on returns 404/410 one morning because the provider pulled it.
• Capability drift. A model claims tool support in its catalog but fails the actual tool call, or silently drops structured-output mode.

Any one of these turns a working agent into a stream of failed runs, and if your framework has a silent paid fallback, into a surprise bill instead.

How Ficelle absorbs it

Ficelle treats every provider as unreliable by default and routes around failure:

• Cooldowns. A model that errors or rate-limits is put on a timed cooldown so traffic moves to the next candidate instead of hammering a dead endpoint.
• Quarantine. Persistent failures, such as a retired model returning 404/410, or a model the runtime reports as billed, are quarantined out of the pool entirely, not retried into the ground.
• Benchmarked capability. Tool-call, JSON, and vision compliance are measured, not trusted from the catalog. A model that claims a capability but fails the check doesn’t get that work.
• Safe fallback. When the top model is unavailable, Ficelle reroutes to the next working one, within strict-zero rules, so fallback never silently means “paid”.

Why this is the product

Free models are a commodity; they exist on many providers at once. The hard part, and the part worth paying for, is the reliability layer: knowing which model works right now, routing to it, and rerouting the instant it breaks, all while keeping spend explicit and your keys local. That layer is Ficelle.