Introduction

TokenFlux has three products on one account and one balance:

• TokenFlux Agent does work for you. Each agent runs code in a private sandbox, uses tools and the web, and keeps files in its own workspace. You talk to it in the web app or from Telegram, WeChat, Feishu and Slack, and it can run on a schedule.
• TokenFlux Gateway gives your own apps many models through one OpenAI-compatible gateway at https://tokenflux.ai/v1.
• TokenFlux Studio generates images through the API and MCP. Use the Image MCP guide to connect a client and generate an image.

Agents make their model calls through the same gateway, with a dedicated key. One balance pays for all three products, and the usage log shows which key made each call.

AgentsCopy Anchor Link

• Agents: sandbox, workspace, models and billing.
• Sessions: work with an agent in the web app.
• Channels: talk to an agent from chat apps.
• Automations: run a prompt on a schedule.
• Tools and memory: notify, memory, sub-sessions and self-scheduling.

APICopy Anchor Link

The API overview covers the base URL, authentication, streaming and errors. Each endpoint has its own page:

• Models
• Chat Completions
• Responses
• Messages
• Embeddings

Web appCopy Anchor Link

The web app is where you manage your account, balance, API keys and agents. It is available in English and Simplified Chinese.

Create an account at tokenflux.ai/auth/sign-up, then follow the Quickstart.

To connect a desktop or coding client, follow the Cline custom provider setup or the Cherry Studio custom provider setup.

Integrate the gatewayCopy Anchor Link

Use an OpenAI-compatible SDK with https://tokenflux.ai/v1 and an Authorization: Bearer API key. Keep existing request code when the selected model supports the same endpoint and features. Models can differ in tools, image input and reasoning support; check Models & pricing before switching.

Streaming uses server-sent events (SSE). See the API overview for authentication, protocol limits, errors and billing. Usage shows each request's key, model, tokens and cost.

For a first SDK call, follow the Quickstart. For a coding workflow, follow the Codex setup. For help, contact support@tokenflux.ai.