Skip to main content

Documentation Index

Fetch the complete documentation index at: https://apixo.ai/docs/llms.txt

Use this file to discover all available pages before exploring further.

This guide walks you through installing the @apixo/mcp-server package and connecting it to your AI coding tool on Windows, macOS, or Linux. The instructions below cover Cursor, Claude Code, and Codex CLI in detail, plus a generic config that works for any other MCP client.

Prerequisites

The server is published to npm as @apixo/mcp-server. npx -y @apixo/mcp-server always pulls the latest release, so you do not need to reinstall when we update it.

Step 1: Set your API key

Export APIXO_API_KEY as a persistent environment variable so the MCP server can pick it up on every launch.
[Environment]::SetEnvironmentVariable("APIXO_API_KEY", "your_real_key", "User")
After running this, close and reopen PowerShell (or your terminal) so the new variable is loaded.Set APIXO_API_KEY on Windows
You can also set APIXO_API_KEY only inside the MCP client config (see Step 2). Setting it as a user-level environment variable is recommended because it works for all clients on the machine.

Step 2: Register the MCP server

Open or create ~/.cursor/mcp.json (Windows: %USERPROFILE%\.cursor\mcp.json) and add an entry under mcpServers:
{
  "mcpServers": {
    "apixo": {
      "command": "npx",
      "args": ["-y", "@apixo/mcp-server"],
      "env": {
        "APIXO_API_KEY": "your_real_key"
      }
    }
  }
}
The env.APIXO_API_KEY field is optional if you already set the variable in Step 1, but adding it here makes the config self-contained.Then open Cursor’s MCP settings panel and confirm that apixo is listed and reports a healthy status.

Step 3: Verify the registration

A healthy install exposes five tools under the apixo server:
  • apixo_list_models
  • apixo_get_model_schema
  • apixo_generate_task
  • apixo_get_task_status
  • apixo_get_balance
How you confirm this depends on the client:
Reload Cursor (or restart it) and open the MCP / Tools panel. The apixo server should report a healthy status and list the five tools above.
If the server is registered but no tools appear, see Troubleshooting -> No tools listed.

Step 4: Try it from your agent

Fully restart your client (Cursor, Claude Code, Codex, or whichever one you set up) so the MCP server is launched, then run the prompts below. Each one constrains the agent to the apixo tools so you can confirm wiring without other tools interfering. This is the primary use case — let the agent write APIXO calls into a small project.
Use only the apixo MCP tools. Build a minimal HTML + JavaScript playground that
calls the nano-banana-2 model through the APIXO API. Look up the model's
parameters yourself, then give me the full code I can run locally.
Expected: the agent calls apixo_list_models and apixo_get_model_schema to discover nano-banana-2’s parameters, then writes working code against https://api.apixo.ai/api/v1/generateTask/nano-banana-2. Test 1 — agent writes integration code in Codex Test 1 — generated frontend playground

Test 2: Submit a task directly from chat

Useful when you want to sanity-check a model before wiring it into code.
Use only the apixo MCP tools. Generate a 16:9 image with the nano-banana-2 model
of a serene Japanese garden at golden hour. Return the final image URL when done.
Expected: the agent calls apixo_generate_task for nano-banana-2, polls with apixo_get_task_status until the task succeeds, and prints the result URL.

Test 3: Check your account balance

Use only the apixo MCP tools. What is the current balance on my APIXO API key?
Expected: the agent calls apixo_get_balance and reports a numeric balance. Test 3 — balance check in Codex

Next steps

What APIXO MCP can do

Tool reference and capability overview.

Troubleshooting

Fix common install issues across clients and operating systems.

Generation API reference

The endpoint the MCP server submits tasks to.

Browse models

Find the model ID to pass to apixo_generate_task.