# Reminix with coding agents

> Let Claude Code, Codex or Cursor find the scripts in your project, test a draft on Reminix and publish it when you say so.

Claude Code, Codex and Cursor already know your code, scripts included. Connect one to Reminix, and it turns those scripts into tools your whole team can run. It finds them, prepares each one, tests it and publishes it when you say so. Connect it, paste one prompt, and pick a script.

## 1. Connect the agent

Each agent can use Reminix’s MCP server. In Claude Code:

```bash
claude mcp add --transport http reminix https://mcp.reminix.com/mcp
```

Then run `/mcp` in Claude Code to sign in. In Codex:

```bash
codex mcp add reminix --url https://mcp.reminix.com/mcp
codex mcp login reminix
```

In Cursor, add the server to `.cursor/mcp.json` in your project, or to `~/.cursor/mcp.json` for every project:

```json
{
  "mcpServers": {
    "reminix": { "url": "https://mcp.reminix.com/mcp" }
  }
}
```

Cursor asks you to sign in the first time it connects. In each case, your browser opens. Choose one workspace and what the agent may do, then click **Allow**.

An agent with a terminal can use the [CLI](/docs/cli/) instead:

```bash
npm install -g @reminix/cli
reminix login
```

When you sign in, keep the access at **Publish and run scripts**. With `--json`, every command prints exactly the JSON the API returns.

## 2. Paste this prompt

```text
Read https://reminix.com/docs/skill/SKILL.md and follow it. List the scripts in this project that people run by hand, one line each, and wait for me to pick one to publish to Reminix.
```

The prompt has the agent read Reminix’s [agent skill](/docs/agents/). It changes nothing until you pick a script.

## What you see when it works

1. The agent lists the scripts in your project that people run by hand.
2. Once you pick one, it adds a `reminix.json` and a `README.md`, and uploads a draft version. Your team can’t run the draft yet.
3. It test-runs the draft and shows you the result.
4. It asks you to publish that version. Once you do, your team can run the script from a form, the API or their own AI agent.

[Publish a script with your agent](/docs/agent-quickstart/) has each step and the commands the agent runs.
