The AI Subscription
Developers Must Have
Delegate your tasks with open models.
No card needed
Start free in the tool you already use.
Keep Claude Code. Run it on open models.
- 1Create your free account · no cardNo card. Your key is ready straight away, and the dashboard shows these lines with it filled in.
- 2Run this once in your terminal
curl -sL -H "Authorization: Bearer sk-cv-…" "https://api.openadapter.in/api/setup/claude-code?sonnet=Kimi-K2.5&haiku=MiniMax-M2.7&opus=glm-5" | bash
Kimi-K2.5, MiniMax-M2.7 and GLM-5 take its Sonnet, Haiku and Opus slots. Change them any time in your dashboard.
- 3Then use it as usual$ claudemodel: Kimi-K2.5 via OpenAdapter› fix the failing checkout test✓ 38 passed
See what it can make.
Open the chatEvery open model. Pick one,
or let auto pick.
Pass a tier as the model name and we route each request to the right model for it.
Keep what you already use.
Stay on your Claude plan, and point your app or scripts at our API for open models, embeddings, speech and search.
Get an API key
Keep Claude Code as your editor and call Koda from inside it. Koda takes the task and runs it on open models, on your plan.
How to call Koda
Add keys from OpenCode, Z.AI Coding Plan, Anthropic, OpenAI and more. Use them in Koda and through our API. That traffic uses none of your plan’s quota.
Add your keysFrom people on a plan.
Leave a reviewI’ve been using the Max Plan, and it’s honestly one of the best AI subscriptions I’ve bought. You get access to 70+ models, including the latest ones, at a much lower price than platforms like Claude. I’m a pretty heavy user, and I still haven’t used even 50% of my monthly limit. Unless you really need the extra speed, I don’t see much reason to pay more elsewhere. OpenAdapter offers amazing value for the price. Highly recommend it!
Speed is great, limits aren’t bad. Context little shorter but hey I get the marketing technique, it’s for sure a interesting one that nobody else deployed.
Best when it comes to Pricing nothing comes near this actually. Good thing I left opencode for this. KODA is a nice additive.
Fast token responses, the models are current, and the limits are generous enough I’ve never hit any of them or really had to worry about hitting them.
I came from opencode and I think its the best decision I have ever done in a while.
How much will you use it?
Drag to your usual day. We suggest the smallest plan that keeps you comfortably inside every limit.
Claude is one company’s models. OpenAdapter is one subscription to open models from many labs, like GLM, Kimi, DeepSeek and Qwen, with a chat, a coding agent and an API on the same plan. The models do this work well, but they aren’t Claude, and we’d rather you know that up front.
Yes. Keep Claude for chat and use our API for your own app, or call Koda from inside Claude Code and hand it the tasks you’d rather run on open models. Nothing here asks you to switch.
Each call spends request credits by model weight: a standard model costs 1, light models 0.5 to 0.75, flagships 2 to 3 and the top tier 5. Embeddings and web tools cost less. Paid plans get a budget every 5 hours, every week and every month, Free gets 50 a day, and each budget refills as its window rolls forward.
There are two meters. Go faster than your per-minute pace and calls slow down for a few seconds. Use up the request credits in a window and they come back as older requests age out. Your dashboard shows what is left, and you can move to a bigger plan any time.
Yes. Set two environment variables, or run the one-line setup from your dashboard. You keep Claude Code itself; the model behind it is an open one, like GLM, Kimi or DeepSeek, not Claude. Cursor, Cline, Roo Code, Kilo Code and OpenCode work the same way.
Every model sits in one of four tiers: Freemium, Lite, Pro and Max. Free reaches the first three. Go reaches every tier across 11 efficient models hosted on 0G. Lite, Pro and Max include every model in the live catalog.
Yes. In Settings, add keys from OpenCode, Z.AI Coding Plan, Anthropic, OpenAI and others. They work in Koda and through our API, and those requests use none of your plan’s quota.
Our coding agent for the terminal. Install it with the one-line setup from your dashboard and it runs on your plan. You can also call it from inside Claude Code, or steer a session from your phone with Koda Remote.