Terms.

The contract between you and Ainfera. Short on purpose — the audit is the contract; this is the boilerplate.

What this is.

By signing up with GitHub or routing a call to api.ainfera.ai, you agree to these terms. They cover the boring legal scaffolding around our actual contract — which is: we route your inference within your caps, settle the bill honestly, and post every decision to a public audit chain that we can't rewrite. If we deviate from that, you have a real grievance. These terms describe the rest.

Your account.

You sign in with GitHub. You're responsible for keeping the GitHub account secure and for actions taken through any API key you generate. Keys are scoped per agent; rotate them at the cadence your secret store enforces. We do not store your GitHub password and we cannot access your GitHub repositories.

You must be at least 18 and legally capable of entering into this agreement in your jurisdiction.

What we provide.

An inference router: you call our API, we choose a model within your caps, send the call to that model's provider, return the response, and settle the bill. We also operate the dashboard, the audit chain, the MCP server, and the public read endpoints.

We use commercially reasonable efforts to keep all of it up. Real numbers live on /status.

Caps and quiet downgrades.

We will not violate a cap you set. If no model fits, we return the error below. We will never quietly downgrade quality, ignore a budget limit, or route a call you said don't route. If we do, that's a breach of this contract — verifiable from the on-chain decision record.

HTTP/1.1 409 Conflict
Content-Type: application/json

{
  "error": "no_eligible_model",
  "reason": "all candidates excluded by caps",
  "caps_applied": { "budget": 0.003, "latency_ms": 400 }
}

Acceptable use.

You may not use Ainfera to: violate law, generate CSAM, deceive humans into believing they're talking to a human in regulated contexts, perform mass surveillance, or compete with a model's own ToS that you've separately agreed to (we propagate provider ToS to your traffic).

We may suspend an account for suspected violation, with the routing decision and reasoning posted on chain. You may appeal in writing to hello@ainfera.ai.

The chain is the chain.

Every settled routing decision lands on the Ainfera audit chain. By using the service, you accept that the chain is immutable and that the metadata posted there (decision hash, prompt hash, response hash, cost, timestamp) cannot be retracted, even on account closure or deletion request.

This is a feature, not a bug. The chain is what makes Ainfera's decisions verifiable.

Payment.

Free during beta. Payment turns on at GA, with at least 30 days' notice in the changelog and by email. At GA, you pay direct inference cost + an 8% flat margin. No platform fee, seat, or volume tier. See /pricing for the worked math.

Invoices are issued monthly. Failure to pay results in agent pause (not silent throttling) with notice.

Your content. Our software.

You retain all rights to your prompts, your responses, your workflows, your templates. We claim no ownership of any of it. We use it only to operate the service for you, as described in /privacy.

The Ainfera software, the audit chain protocol, the mark, the wordmark, the documentation are ours. You may not redistribute or sublicense them.

Warranties.

The service is provided "as is" during beta. At GA, we will warrant operational reliability targets explicitly in a separately published SLA. Audit immutability is the one warranty we make today: we will not retract or modify on-chain records, even if you ask us to.

Liability.

To the maximum extent permitted by law, our aggregate liability is capped at fees paid in the trailing 12 months, or $100 USD if no fees have been paid yet. We are not liable for indirect, consequential, or incidental damages — including agent overspend if you didn't set a cap.

This isn't us trying to wriggle out. It's the standard infrastructure liability shape. We do not handle your billing system and cannot underwrite your agent's actions.

# in plain terms:
liability_cap = max(fees_paid_trailing_12mo, $100)
indirect_damages = excluded
agent_overspend_without_cap = your_responsibility

Termination.

You may close your account at any time from the dashboard. We process the request immediately; your data is removed per the retention rules in /privacy. On-chain hashes persist, by design.

We may terminate accounts for breach (notably acceptable-use violations or unpaid invoices), with notice.

Disputes.

These terms are governed by the law of the State of Delaware. Disputes are resolved by binding arbitration before JAMS in San Francisco, except that either party may seek injunctive relief in court. You may opt out of arbitration by emailing legal@ainfera.ai within 30 days of signing up.

How these terms change.

Material changes get at least 30 days' notice in the changelog and by email. You may close your account before they take effect if you don't agree. Non-material changes (typos, link updates) are immediate and noted in the version history.

Every version of these terms is signed by the team and committed to the audit chain. See the meta column above for the diff link.