Subscription plans
Plans and billing

Subscription plans

Subscription plans & what they include#

This page explains how Rayu bills you, what each plan unlocks, and where to see your live numbers. Plans, prices, and rates can change over time, so treat the tables here as a guide and check your account (see Checking your plan & balance) for what applies to you today.

Two ways to use Rayu#

You can run Rayu in two modes, in either the CLI or the VS Code extension:

Rayu account (hosted)Bring your own key (BYOK)
What it isSign in with a Rayu account (/login). Model requests go to a curated set of models hosted by Rayu.You connect your own provider API keys (/connect). Requests go straight from your machine to the provider.
BillingIn credits, drawn from your plan's monthly allowance (plus an optional pay-as-you-go wallet). BYOK traffic is not metered by Rayu — you pay your provider.Nothing to Rayu.
Works onAny plan; hosted models start at Basic.Every plan, including Free — Free is BYOK-first.
ExtrasDaily-turn caps don't apply on paid plans; features like Telegram, Orchestrator mode, and image/video generation unlock with the plan.The full tool set of the agent itself (files, shell, MCP, skills) is yours on every plan.

You can switch between the two at any time with /model — nothing about your account prevents you from using your own keys.

Plan comparison#

PlanPriceCredits per monthDaily turnsPacing: weekly budgetPacing: per 5-hour windowTop-ups
Free$0none (BYOK only)50 turns/daynot applicablenot applicableno
Basic$3/mo15,000unlimited3,7505% of the weekly budgetyes
Pro$10/mo50,000unlimited12,5005% of the weekly budgetyes
Plus$20/mo115,000unlimited28,7505% of the weekly budgetyes
Max$50/mo300,000unlimited75,0005% of the weekly budgetyes
Enterprisecoming soon — contact salesnegotiatednoneper contractper contractper contract

How to read the pacing columns:

  • Your monthly allowance is not usable all at once. Rayu also applies a rolling weekly budget and a per-5-hour-session cap (5% of the weekly budget — e.g. Pro releases at most ~625 credits per 5-hour window from a 12,500 weekly budget). When a window is exhausted, requests wait until it frees; your credits are not lost.
  • Free's limit is different in kind: a 50-turn-per-day cap that applies to hosted turns (your BYOK requests never count toward it), and no hosted models at all.
  • Concurrency (how many requests stream at once) and the session-window length (default 300 minutes) are service-wide settings, not per-plan numbers.

Feature access by plan#

Rayu unlocks a number of features with your plan. Current defaults:

Gated featureWhat it isFreeBasicProPlusMax
Telegram botDrive sessions from Telegramoffonononon
Orchestrator modeMulti-agent delegationoffonononon
Delegated-worker model overrideModel choice for delegated background workersoffonononon
Model per subagentRoute subagents to a different model/provideroffonononon
Image generationNative image tooloff*onononon
Video generationNative video tooloff*onononon
Multiple API keysSeveral keys per provider with failoveroffonononon
API accessProgrammatic keys for hosted models (OpenAI- and Anthropic-compatible)offoffononon
OCR (document scanning)Scan documents with a hosted modeloffonononon
Audio (STT / TTS)Speech-to-text and text-to-speech with a hosted modeloffonononon

Notes:

  • * Image and video can also be paid for from a funded pay-as-you-go balance even when your plan does not grant them — so you can pay per use without subscribing, where top-ups are available. Every other gated feature needs the plan.
  • Basic deliberately keeps API access off (every other feature is on with Basic).
  • Enterprise (coming soon) turns all features on with negotiated limits.
  • Gating is soft in the UI (a locked feature shows you an "upgrade required" note instead of disappearing) but enforced for anything that runs on Rayu's hosted models.

Hosted models and example credit rates#

With a Rayu account on Basic, Pro, Plus, or Max, you get the hosted model catalog — the models Rayu hosts for you, billed in credits. The Free plan has no hosted models (BYOK only).

Credit cost is computed per token bucket: credits = (input·rIn + output·rOut + cacheRead·rRead + cacheWrite·rWrite) / 1,000,000 where each r is that model's credits-per-1M rate. See How token credits are charged for the full mechanics.

Every model below is available on Basic and above. Rates are in credits per 1M tokens:

Model (code)InputOutputCache readCache writeContext
DeepSeek V4 Flash (deepseek-v4-flash)330330333301M
DeepSeek V4 Pro (deepseek-v4-pro)1,0001,0001001,0001M
LongCat 2.0 (longcat-2)5002,000105001M
GLM-5.2 (Ollama Cloud) (glm-5.2)2,5002,5002502,5001M
Kimi K2.7 Code (Ollama Cloud) (kimi-k2.7)2,5002,5002502,500256K
MiniMax M3 (Ollama Cloud) (minimax-m3)2,5002,5002502,500128K
Llama 4 (Ollama Cloud) (llama-4)1,0001,0001001,000128K
GPT-OSS 120B (Ollama Cloud) (gpt-oss-120b)75075075750128K
Qwen3.5 397B (Ollama Cloud) (qwen3.5-397b)75075075750256K
Qwen3.5 122B (Ollama Cloud) (qwen3.5-122b)75075075750256K

A worked example: one Pro-plan turn on deepseek-v4-pro using 8,000 input + 2,000 output tokens with no caching costs (8,000·1,000 + 2,000·1,000) / 1M = 10 credits. Cache hits are the big saver in long sessions — on DeepSeek V4 Pro a cached input token costs a tenth of an uncached one, and agentic loops hit cache for most of the prompt.

Note: these are current default rates. Plans, prices, model availability, and every rate can change. For live values check /billing or /usage in the CLI/extension, or rayucode.com/plans.

Which providers exist per path#

PathProviders
Hosted (Rayu account)The curated catalog above — models hosted by Rayu. The exact set follows what Rayu offers at the time.
BYOK (/connect)Any of the ~30 provider presets regardless of plan — Anthropic, OpenAI, NVIDIA, DeepSeek, Kimi, Gemini, Bedrock, Azure, Vertex, Copilot, Kiro, OpenRouter, Fireworks, Together, Cerebras, DeepInfra, HuggingFace, xAI, Groq, MiniMax, LongCat, local OpenAI-compatible servers, and custom endpoints. See Providers.

A single session can mix them: main agent on a hosted model, a subagent on your own provider key (paid plans include per-subagent model routing).

Top-ups and pay-as-you-go#

Beyond your plan allowance, you can buy pay-as-you-go (PAYG) credits:

  • In the terminal or the extension: /topup [credits] (alias /buy-credits) — shows the live price quote and a payment flow (card or QR).
  • In the dashboard: rayucode.com → billing.
  • Top-up availability and the credit-per-dollar rate can change; if purchasing is unavailable, /topup tells you. The minimum top-up is $1. Paid plans (Basic and up) include top-up spending.

How money is spent, per request:

  1. Plan credits first. The current period's subscription allowance is drawn down to zero.
  2. Then your PAYG wallet. Whatever the plan could not cover comes from your permanent top-up balance. (This split happens within a single request, so the last few credits of a period are never wasted.)
  3. Then refusal. If neither balance covers the cost, the request is denied rather than silently over-billed — top up, wait for the window to free, or switch to BYOK.

Two important properties:

  • The two balances never merge. Plan credits reset with your billing period; PAYG credits are permanent. They are tracked separately and drawn in the order above.
  • Pacing still applies. The weekly and 5-hour windows gate the rate of hosted spending even when your balances are healthy; that is a pace, not a budget. See Credits & limits.

Checking your plan & balance#

Command / pageWhat you see
/login (and /logout)Sign in or out of the Rayu account that carries your plan.
/billingYour current plan, and a link to upgrade on the web.
/usagePlan allowance plus hosted-model usage: what you have spent and what is left.
/topupBuy PAYG credits at the live rate.
/statsYour local session usage statistics and activity.
/insightsA generated report analyzing your Rayu sessions.
Dashboardrayucode.com — plan, invoices, credit balance, and plan changes in the browser.

Billing commands require a Rayu account sign-in. Upgrades and downgrades (prorated) are handled on the dashboard; the plan change takes effect for the CLI and extension automatically — no reinstall.

See also#