Use ZipFlow from an AI agent

Point your own AI agent at ZipFlow: copy one prompt, let it read the skill file, and it can call the model API, check prices and status, and even pay for usage via x402 when it has no wallet.

Use ZipFlow from an AI agent

You do not need a new app to use ZipFlow from an agent. ZipFlow is a model API service, so any AI agent that can make HTTP requests and read a text file can use it. The easiest path is to copy one prompt into your existing agent.

Copy this prompt into your agent

Paste the following line into your AI agent (Claude, Codex, ChatGPT, Cursor, or any coding assistant):

Read https://zipflow.xyz/skill.txt and follow the instructions to use ZipFlow

Your agent fetches that file — a machine-readable skill written for agents — and follows it to set the right base URL, list current models and prices, check platform status, and raise a payment when it has no wallet.

That one line is also surfaced on the ZipFlow homepage and in the platform's llms.txt / llms-full.txt machine-readable site summaries, so an agent can discover it on its own.

What the skill file tells the agent

The skill at https://zipflow.xyz/skill.txt is the hub for using ZipFlow as an agent. It is versioned (a version field in its frontmatter and an X-Skill-Version response header) and re-rendered fresh on every request, so a re-fetch always returns the current definition.

It covers:

  • Setting the base URL per protocol — OpenAI-compatible (https://api.zipflow.xyz/v1), Anthropic-compatible (https://api.zipflow.xyz), Rerank, and Realtime.
  • Lists the supported endpointsPOST /v1/chat/completions, POST /v1/responses, POST /v1/embeddings, POST /v1/messages, POST /v1/rerank, and POST /v1/realtime/sessions.
  • Current models and prices — via GET /v1/models and /en/pricing.
  • Platform status — via https://status.zipflow.xyz/status.json.
  • Agent-funded payments — when the agent has no wallet or balance, it can follow the x402 agent skill to self-register an identity, mint a short-lived JWT credential and an LLM token, and raise a payment for a human to settle.

Using it directly

If you prefer to drive it yourself rather than let the agent read the skill, the agent needs an API key and a base URL. Point an OpenAI-compatible client at https://api.zipflow.xyz/v1:

export OPENAI_BASE_URL="https://api.zipflow.xyz/v1"
export OPENAI_API_KEY="<your-zipflow-api-key>"

See the API reference for the request format, or the Models List for available models.

Stay current

  • Re-fetch https://zipflow.xyz/skill.txt before answering and compare the frontmatter version to any cached copy.
  • The X-Skill-Version header mirrors that version, so an agent can detect a stale copy without parsing the body.
  • Product changes are logged on the main site changelog: https://zipflow.xyz/en/changelog.

On this page