Chat with any connected LLM. Pay-as-you-go in tokens.

Pick a provider, pick a model (flagship / fast / budget), send a message. Cost deducted from your Token Wallet in real time. Full conversation history.

Get a Token Wallet See providers

What you get

OpenAI-compatible

Most providers (OpenAI, DeepSeek, xAI, Groq, OpenRouter, Perplexity, Cohere) use the standard /v1/chat/completions endpoint.

Anthropic + Google + MiniMax

Anthropic uses x-api-key + system field. Google uses contents.parts. MiniMax uses /v1/text/chatcompletion_v2.

Model picker

Flagship / fast / budget tiers per provider. Switch with one click. No model retraining, no cache invalidation.

Token-metered

Each request estimates max cost first. If your Token Wallet can't cover, you're told. No surprise overage.

Audit log

Every chat is logged with prompt, completion, tokens, cost, and provider. Stored in user_llm_usage_log.

Cost transparency

You see $/request and total $ in the response. No "contact sales for usage".

In practice

AI inside · Devon · using MiniMax for quick drafts

Devon uses his Token Wallet to chat with MiniMax-Text-01 (fast tier). He sends 50 messages in a day, each ~$0.0003. His wallet deducts $0.015 total. No subscription, no minimum.

$0.0003

Avg cost per message

At a glance

LLM providers

Tiers (flagship/fast/budget)

200

Tokens per $1

~2s

First-token latency