Tutorial5 min read

Connecting Dolphin Anty to Ornold: Fast MCP Setup

From enabling the debug port to listing connected profiles, this walkthrough gets your first live browser batch online in a few minutes.

Mar 28, 2026

One Prompt — Your AI Agent Connects to Dolphin Anty

You don't need to find ports, copy API tokens between docs, or pick flags. Paste one prompt with your tokens. Your AI agent (Claude Code, Cursor, Codex, Windsurf, Cline) reads the public install guide on GitHub, asks you 1-2 questions, then connects to Dolphin Anty automatically.

How it works: Ornold publishes a public agent guide at github.com/ornold-ai/ornold-mcp. Any AI agent can read `AGENTS.md` and figure out how to install correctly for Dolphin Anty.

Step 1 — Prepare Dolphin Anty

Open the Dolphin Anty desktop app
Settings → enable **Local API** (default port `3001`)
Sign in to dolphin-anty.com, open API → copy your **API token**
Make sure you have at least one profile created

Step 2 — Get Your Ornold Token

Sign up at mcp.ornold.com
Open the Tokens page
Click Create token, copy the value (starts with `orn_sk_`)

Step 3 — Paste This Prompt into Your AI Agent

Open your agent (Claude Code, Cursor in Agent mode, Codex, Windsurf, Cline) and paste:

Install Ornold MCP for Dolphin Anty automation. I'm new to this, so guide me.

1. Fetch and read the official install guide:
   https://raw.githubusercontent.com/ornold-ai/ornold-mcp/main/AGENTS.md

2. Detect which AI agent I'm running in (Claude Code, Cursor, Codex, Windsurf, Cline) and use the correct install method from AGENTS.md.

3. Ask me:
   a) What port is Dolphin Anty's Local API using? (Default 3001 — I'll tell you if I changed it.)
   b) Which interaction mode do I want? Briefly explain "vision" (AI sees the page like a human, recommended for antidetect) vs "dom" (faster but easier to detect). Recommend vision.

4. Install with these flags: --dolphin-port <my port> --dolphin-token MY_DOLPHIN_API_TOKEN --mode <my mode> and my Ornold token.
   - My Ornold token: YOUR_ORNOLD_TOKEN
   - My Dolphin API token: YOUR_DOLPHIN_API_TOKEN

5. If I'm using Claude Code, also install the AI skill: npx mcp-ornold-browser --install-skill

6. Verify by calling browser_list. Tell me how many Dolphin profiles are detected.

If anything fails, explain in plain language.

Replace `YOUR_ORNOLD_TOKEN` and `YOUR_DOLPHIN_API_TOKEN` with your real tokens. The agent will fetch the guide, ask 1-2 questions, install, and confirm.

Why Two Tokens?

**Ornold token (`orn_sk_...`)** — authenticates you with the Ornold MCP server (billing, AI features).
**Dolphin API token** — lets Ornold talk to Dolphin Anty's API (start/stop profiles, manage sessions).

Both are required. Your Dolphin token stays on your machine — it's used locally to talk to your Dolphin app.

Why Vision Mode for Dolphin?

Dolphin profiles have unique fingerprints, but if your AI clicks via DOM selectors and dispatches synthetic events, anti-fraud detects the automation pattern instantly — burning the profile and defeating antidetect entirely.

`--mode vision` makes the AI take screenshots, identify elements visually, then click by coordinates with human-like Bezier mouse curves and gaussian typing delays. Vision mode is also required for flow recording (record once on one profile, replay across many).

Step 4 — Verify

Ask the agent:

List my Dolphin Anty profiles using dolphin_get_profiles

You should see your profiles. Try a real action:

Start the first profile and navigate to google.com

Already Know What You Want? Skip the Questions

Claude Code

claude mcp add --transport stdio ornold-browser -- npx mcp-ornold-browser --token YOUR_ORNOLD_TOKEN --dolphin-port 3001 --dolphin-token YOUR_DOLPHIN_API_TOKEN --mode vision
npx mcp-ornold-browser --install-skill

Cursor (~/.cursor/mcp.json)

{
  "mcpServers": {
    "ornold-browser": {
      "command": "npx",
      "args": ["mcp-ornold-browser", "--token", "YOUR_ORNOLD_TOKEN", "--dolphin-port", "3001", "--dolphin-token", "YOUR_DOLPHIN_API_TOKEN", "--mode", "vision"]
    }
  }
}

Codex (~/.codex/config.toml)

[mcp_servers.ornold-browser]
command = "npx"
args = ["-y", "mcp-ornold-browser", "--token", "YOUR_ORNOLD_TOKEN", "--dolphin-port", "3001", "--dolphin-token", "YOUR_DOLPHIN_API_TOKEN", "--mode", "vision"]

Public Resources

**GitHub** — github.com/ornold-ai/ornold-mcp — public agent install guide.
**npm** — mcp-ornold-browser
**Dashboard** — mcp.ornold.com

Troubleshooting

**`browser_list` empty / `dolphin_get_profiles` fails** — Dolphin Anty must be running with Local API enabled (Settings → Local API → on).
**`Connection refused` on port 3001** — port may be different. Check Settings → Local API for your real port and update `--dolphin-port`.
**`401 Unauthorized` from Dolphin** — your Dolphin API token is wrong or expired. Re-copy from dolphin-anty.com → API.
**Ornold token invalid** — re-copy from mcp.ornold.com Tokens page.
**npx: command not found** — install Node.js 18+ from nodejs.org.

What's Next?

Parallel browser automation — run actions across many Dolphin profiles at once
CAPTCHA solving guide — reCAPTCHA, hCaptcha, Turnstile
Vision AI vs CSS selectors — why this matters for antidetect
Scaling account registration to 50 profiles

Explainer