# Getting started

A remote MCP server that lets a coding agent audit a live page, plan conversion changes from a library of evidence-graded tactics observed on top SaaS sites, implement them in your codebase and verify them after release.

You talk to your coding agent (Claude Code, Cursor, Codex, Windsurf or any MCP client). The agent calls Liftwire tools, makes the changes in your codebase, and checks them on the live page. Nothing is installed on your site.

## 1. Get an API key

Request a free key at [https://liftwire.io/mcp](https://liftwire.io/mcp). We email it once and store only a hash of it, so keep the email. The free plan covers one site; see [Limits and pricing](/docs/limits) for the paid plans.

## 2. Connect your agent

The MCP endpoint is:

```
https://liftwire.io/mcp/http
```

It speaks Streamable HTTP and takes your key as a Bearer token.

**Claude Code** (Run in a terminal)

```bash
claude mcp add --transport http liftwire https://liftwire.io/mcp/http --header "Authorization: Bearer <your-api-key>"
```

**Cursor** (~/.cursor/mcp.json (or .cursor/mcp.json in a project))

```json
{
  "mcpServers": {
    "liftwire": {
      "url": "https://liftwire.io/mcp/http",
      "headers": {
        "Authorization": "Bearer <your-api-key>"
      }
    }
  }
}
```

**Codex** (~/.codex/config.toml, with LIFTWIRE_API_KEY exported in your shell)

```toml
[mcp_servers.liftwire]
url = "https://liftwire.io/mcp/http"
bearer_token_env_var = "LIFTWIRE_API_KEY"

# Then add this line to your shell profile (not to this file):
# export LIFTWIRE_API_KEY="<your-api-key>"
```

**Windsurf** (~/.codeium/windsurf/mcp_config.json (Windsurf Settings > Cascade > MCP servers > View raw config))

```json
{
  "mcpServers": {
    "liftwire": {
      "serverUrl": "https://liftwire.io/mcp/http",
      "headers": {
        "Authorization": "Bearer <your-api-key>"
      }
    }
  }
}
```

The setup page at [https://liftwire.io/mcp](https://liftwire.io/mcp) has the same snippets with copy buttons. To teach your agent the whole loop as well, install the free [agent skill and plugin](/docs/skill).

## 3. Run a first audit

Ask your agent, in your project:

> Audit the pricing page at https://your-site.example/pricing and plan the three most valuable changes.

The agent will call `site_memory`, then `audit_page`, poll `get_audit`, and call `recommend`. Audits run in the background and take a minute or two; the tools tell the agent how long to wait between polls. [The loop](/docs/the-loop) walks through every step.

## Before you rely on it

The tactic library is young. Every response tells your agent how many tactics are published (`tactic_count`), and every tactic carries an [evidence grade](/docs/evidence-grades). The [sell-readiness page](/docs/sell-readiness) shows the thresholds we hold ourselves to and where the corpus stands against them.