Stop juggling lab accounts. One key.
Qwen, DeepSeek, Kimi, GLM and more through one OpenAI-compatible endpoint, with your coding agent set up in five minutes and a real person to help.
# Run your coding agent on an open model (example) export ANTHROPIC_BASE_URL="https://api.meruva.ai/anthropic" export ANTHROPIC_AUTH_TOKEN="MERUVA_API_KEY" export ANTHROPIC_MODEL="kimi-k2" claude # Or any OpenAI SDK base_url = "https://api.meruva.ai/v1" model = "deepseek-chat" # change one string # Example endpoints and model names; final values in the docs.
What gets in the way, and what we do about it.
A console per lab
Separate accounts, top-ups and dashboards for every model family.
Different formats
Every provider's endpoint and parameters differ slightly.
Agent setup by trial and error
Agent config, base URLs and model names take an afternoon.
Surprise bills
Hard to see what each call or session cost.
Cache pricing is a mystery
Cache reads can cut agent costs, if you know how they're billed.
Lock-in
Switching providers means rewriting integrations.
Three steps to your first call.
Coming soon You cannot sign up, top up or get a key yet.
Sign up and top up by card through Stripe, when we open.
Create a key in the dashboard.
Connect your tools. Paste the key into your app or coding agent, then watch usage and spend.
Open models only.
| Family | Good for (examples) |
|---|---|
| Qwen | General coding and chat, long context |
| DeepSeek | Reasoning and low-cost bulk work |
| Kimi | Agentic coding and tool use |
| GLM | Writing, review and structured work |
| More | Added as labs release them |
Model strengths are examples; the live list will be on the models page. Prices on Pricing, marked as examples until final.
Meruva Agent Orchestrator.
Open-source agents that run on these models, with human approval gates at plan, design and merge. Being open-sourced, and not available yet.
Your first call in five minutes.
Coming soonKeep it in your jurisdiction.
The same models, deployed on your own infrastructure.
See Sovereign