Try it
Free · no key- Model
- Why
- Size
Previews are free and nothing is stored. With your own key, each decision is $0.0002.
Send Raeto the request you're about to make and it tells you which model to use: cheap ones for simple asks, strong ones for hard work. You call the model with your own provider key, so nothing about your setup changes.
Previews are free and nothing is stored. With your own key, each decision is $0.0002.
Your keys, agents, usage and billing in one place.
| Model | Calls | Spent |
|---|
Auto-routed calls and routing decisions, newest first.
| Time | Key | Model | Status | Cost | Saved |
|---|
Savings from auto routing, prompt caching and the loop breaker show here once you’re making calls.
Leave any time. It’s always under your wallet, at the top right: Guided setup.
| Agent | Status | Balance | Today / cap | Last active | Priors |
|---|
See what an agent has actually borrowed and repaid before you or your agents pay it. Records come from Priors, the credit record for agents on Robinhood Chain. Free.
Name the task each call belongs to and Raeto adds them up: what one task costs you, typically and at worst, and what each task type and agent costs.
| Type | Tasks | Typical | Worst | Budget per task | Stops |
|---|
| Ended | Task | Agent | Calls | Cost | How it ended |
|---|
Raeto tells your server when a task finishes, a task or a yield-mode key hits its budget, the loop breaker stops an agent, or your balance runs low. Each delivery is signed with your endpoint's secret. No prompts are ever sent.
x-raeto-signature header with it: HMAC-SHA256 of t.body.
| Endpoint | Last delivery |
|---|
A span for every model call, sent to your tracing backend (Honeycomb, Grafana, Datadog, Langfuse or any OTLP/HTTP collector). No prompts or answers.
Every model you can call through Raeto, priced at your rate. Use raeto/auto and Raeto picks the cheapest one that can do the job.
| Model | Context | Input / 1M | Output / 1M |
|---|
Your AI app gets 400+ models, every tool in Tools, Raeto Pay and an agent wallet, all paid from this balance inside the limits you set.
Wrap the client you already use. If Raeto is ever slow or unreachable, the request goes to your usual model, so it can't break your app.
Set a key to spend from vault yield and it spends, each day, at most what your vault deposits paid. A light agent runs on its own yield without ever being topped up.
| Key | Spends from | Typical call | Calls a day on yield | Deposit it needs | Overdraft a day |
|---|
Web search, onchain data and APIs from other builders. Your agents find them and pay per call from your balance. Or list your own API and get paid by every agent on Raeto.
Web search, onchain data, token research and APIs from other builders. Agents call them over MCP or REST, paid from your balance, inside your key limits. Charged only when a call works.
Connect an agentBrowse belowList an endpoint behind Raeto Pay. Agents find it with find_tools and pay your price per call. Raeto keeps 1.5%.
| Tool | What it does | Price |
|---|
Sell your API to every agent on Raeto. Agents find it with find_tools, call it with use_tool and pay your price per call from their balance. Raeto keeps 1.5%. How it works
| Tool | Status | Price | Calls | Success | Earned |
|---|
New and changed tools wait for review, usually within a day. A tool that fails 5 calls in a row is hidden until you fix it. Agents get their money back on any failed call, so failures cost you nothing but your success rate.
model to raeto/auto. Nothing else changes.Save mode (open models at half price through Surplus) is on the Overview, and per key in each key’s settings.
Your calls that name a model, priced again the way Raeto would have run them. Same tokens, same prices, nothing estimated in your favour.
| If you… | Cost | Saves |
|---|
Try raeto/auto on your real traffic before switching. Your calls don't change; Raeto pays for the comparison.
| Key | Sampled | Auto matched | Your model | With auto |
|---|
Across all of Raeto, last 14 days: how often each model's answers had to be retried, per kind of step. Models marked “skipped first” aren't picked first for that kind of step.
| Kind of step | Model | Calls | Retried | Bad tool calls | Avg cost |
|---|
Turn it off per key in the key’s settings (Keys tab), or for one call with the header X-Raeto-Cache: off.
| Time | Agent | Key | What happened | Repeats |
|---|
Set it to warnings only, or off, in each key’s settings. A single call can opt out with the header X-Raeto-Loop-Guard: off.
Each decision uses $0.0002 of your balance. Funds never expire.
Turn balance back into USDG in your wallet. Two confirmations: your balance becomes rCREDIT, then rCREDIT becomes USDG.
What agents paid your API, after Raeto's fee. Take it out as USDG on Robinhood Chain to this account's wallet.
A running total of what this wallet has spent on AI, for points and rewards programs that ask for it. Nothing else about your account: never prompts, keys or balances.
Keep rCREDIT in your own wallet and Raeto tops up this balance from it, a little at a time, up to a daily limit you set.
For teams running real volume, like agent platforms. Tell us what you're building and how many requests you route.
Message @RaetoCC on X Already a partner? Partner sign-in