Rayu Code is an AI coding agent for your terminal and VS Code. Put claude-opus-5.5 in charge with /model, hand the edits to DeepSeek-V4.1-flash:lastest with /subagent_model, and pay from one credit balance, from $1.99 a month. Or bring your own keys for 30+ providers.
Pick the size of the job and see what each hosted model would charge before you run it.
DeepSeek-V4.1-flash:lastest1M context · 500 · 1,500 per 1M26credits
Qwen3.8-flash1M context · 500 · 1,500 per 1M26credits
Minimax-m3.1-flash1M context · 500 · 1,500 per 1M26creditsMinimax-M31M context · 2,500 · 6,000 per 1M124creditsQwen3.8-max1M context · 6,000 · 15,000 per 1M300creditsOne command on macOS and Linux, or PowerShell on Windows. The VS Code extension runs the same engine with the same account.
Run rayu and choose Sign in. Your browser opens, you approve with Google, and the terminal finishes on its own. No API keys to copy.
Set the main model with /model and the subagents with /subagent_model. Every hosted model draws from the same credits.
The agent edits, runs your tests and shows the diff before anything lands. Check what it cost with /usage.
Orchestratorclaude-opus-5.5ledger.ts but settle.ts never reads them back, so a refunded top-up still counts toward the balance until the nightly job reruns.bun run db:migrate?DenyAllow onceThe same agent, account and credits on every surface.
Run Rayu in any terminal or inside VS Code. Both use one engine and one account.
Open the agent in a browser tab and work on your projects without installing anything.
A desktop app for macOS, Windows and Linux, with the same models and credits.
Plan credits renew each month. Top-up credits never expire and are used after your plan credits. Pay by card, or by QR with Bakong KHQR in Cambodia.
10,000 credits a month.
40,000 credits a month.
100,000 credits a month.
240,000 credits a month.
540,000 credits a month.
Full details and checkout on the plans page.
Signed in on a paid plan, you get Rayu’s hosted catalog: claude-haiku-5.5, claude-opus-5.5, claude-sonnet-5-5, DeepSeek-V4.1-flash:lastest, Long Cat 2.5, Mistral-large-4, Qwen3.8-flash, Minimax-m3.1-flash, Kimi-k3.1-pro, kimi-k2.8-for-coding, Long Cat 2.0, claude-opus-4.8, GLM5.3 flash, GLM-5.3, Sonnet 4.6, Minimax-M3, Kimi-k2.7-code, Qwen3.8-max, Opus 4.6, Nemotron-3-Ultra-550b-a55b. The list updates live. With your own keys you can also use Anthropic, OpenAI, Gemini, OpenRouter and about 30 other providers.
Use /model to set the main model, /subagent_model to put subagents on a different one, and /orchestrator to have the main agent plan and delegate the work.
Each hosted model charges credits per million tokens at its own rate, and cache reads cost about a tenth of fresh input. Plan credits are used first, then any top-up credits. Plan credits renew monthly; top-up credits never expire.
Yes, on every plan including Free. Run /connect to add a provider key, or point Rayu at Ollama, LM Studio or vLLM. That traffic goes straight to the provider and uses no Rayu credits.
By card, or by QR with Bakong KHQR in Cambodia. Run /topup in the CLI, or open billing on rayucode.com.
Rayu Code 3.0 brings the agent to the browser, and 4.0 brings a desktop app. Follow the changelog or the Telegram community for release news.
Install the CLI, pick a plan from $1.99, and give every job the model it needs.
