Skip to main content
Everything account-shaped lives in one place: console.dexpaprika.com. Your key, what you have spent this period, the errors your calls returned, the MCP setup for AI clients, and billing. You do not need it to start. The API answers keyless requests, and a key is what raises the allowance and unlocks any-token streaming. See pricing for what each tier includes.
Console dashboard showing four summary cards across the top: current plan, usage this period against the monthly cap, autoscaling state, and next billing cycle. Below them a stacked usage-over-time chart for the last 30 days, and two panels listing top endpoints and top API keys by share of calls.

​
What each section is for

​
Frequently asked

API Keys in the sidebar. A key is generated for your account, and there is exactly one active key at a time.The same screen shows the base URLs your key is valid for. Free keys authenticate against https://api.dexpaprika.com and https://streaming.dexpaprika.com. Pro moves you to https://api-pro.dexpaprika.com and https://streaming-pro.dexpaprika.com.That pairing matters: a key sent to the wrong host answers 403 wrong_host rather than falling back, and the body tells you which host to use.If you have just moved to Pro, read upgrading to Pro before your first call, and 403 Forbidden (wrong host) if you already have the error in front of you.
API Keys screen showing the REST and streaming base URLs with copy buttons, the active key masked with reveal and copy controls, a revoke button, and a quick-start curl example using the Authorization header.
Revoke the current one, then generate a new one. Both buttons are on the API Keys screen.Because there is one active key per account, rotation is a break: the old key stops working the moment you revoke it. Have the new key ready to deploy before you press it, and if the key lives in a hosted environment, remember it is usually stored as a secret rather than in the code.
Put the key in the Authorization header, on its own:
The key is the entire Authorization header value, with nothing in front of it and nothing after it. This is easy to trip over, since almost every other API expects a scheme word first. If your HTTP client only offers a token field that prepends one for you, set a raw header instead.
Billing, at the bottom of the page, with an export control for pulling them out. Until your first payment the section says so rather than sitting empty, so an empty list is not a sign that something failed.The same screen carries the rest of the money: your current plan with the period it runs for and the autoscaling switch.The dashboard shows the next billing cycle at a glance. If it reads no active subscription, you are on a free tier and there is nothing to invoice.
Billing screen showing the current plan card with the monthly allowance, rate limit and billing period plus upgrade buttons, an autoscaling panel with an off switch and a spending-limit slider, three credit pack cards for 5M, 15M and 50M credits, and an invoices section with an export control.
It is how you keep going past the monthly allowance on Dev and Pro. Leave the switch off and the allowance is a hard stop, exactly like a free tier. Turn it on and once the plan is spent, extra credits are bought automatically at $20 per million, up to a spending limit you set on the same screen. You decide the ceiling once and then stop thinking about it.If your account already holds a credit pack, those credits sit alongside the plan allowance and drain with it, and Billing shows what is left. New packs are not on the pricing page; autoscaling is how you add credits now.Autoscaling is a paid-plan control. On a free tier it is visible but inactive.
The dashboard card Usage this period is the short answer: credits used against your monthly cap, with the percentage. Usage breaks it down over time and by endpoint, so you can see which call is doing the spending.One request costs one credit. Batch endpoints charge one credit per item, and each delivered streaming update costs one credit too. See plan your credit usage for how to size a workload before you build it.
On a free tier the cap is hard: the API answers 402 and the body names the next step for your tier. Nothing is charged, and the allowance refills continuously as usage older than 30 days ages out, so there is no date to wait for and the body carries no resets_at.On Pro you can let it keep going. The dashboard calls this autoscaling: with it off the cap behaves like the free tier, with it on the overage is charged per additional block up to the monthly spend limit you set. See error handling for the difference between a 402 and a 429, which is worth getting right in code: a 429 is a per-minute rate limit and is worth retrying, a 402 is the monthly allowance and retrying will not help.
MCP & AI Setup generates the configuration with your key already filled in, for Claude Code, Claude Desktop and any client that takes a JSON config.The hosted server speaks two transports. Streamable HTTP, which is what current clients use:
and SSE, for older clients:
Authentication is the same Authorization header as REST. Full setup per client is in the hosted MCP guide.
PLACEHOLDER. MCP and AI Setup screen with a TL;DR panel showing the server URL and Authorization header, then three cards side by side: a terminal command for Claude Code, a one-click install button for Claude Desktop, and a JSON config block for Cursor, Cline and other clients.
No. Keyless requests work at 30,000 credits per rolling 30 days per IP, with no signup and no card. Registering a free key raises that to 100,000 and lifts the streaming restriction from the showcase list to any token we index.