You can work with your pangoling workspace from Claude, ChatGPT, Cursor, VS Code or Codex. pangoling runs an MCP server (Model Context Protocol, the standard these assistants use to call outside tools), so the assistant can read your keywords, rivals, results and drafts, and save drafts back for you to review.

Everything is set up in Settings, then Connections, then Agent access. Only a workspace admin can create tokens.

Your workspace's MCP URL

Each workspace has its own address:

https://yourbrand.pangol.ing/api/mcp

Copy the exact URL from the Agent access card. An assistant connected to it only ever sees that one workspace.

These two connect with a sign-in, so there is nothing to copy except the URL.

Claude.ai: Settings, then Connectors, then Add custom connector. Paste the MCP URL and sign in.
ChatGPT: Settings, then Apps & Connectors, then Advanced, and turn on Developer mode. Choose Create, paste the MCP URL, and pick OAuth for authentication.

You approve the connection on a pangoling screen. Connected apps are listed on the Agent access card, where you can revoke any of them.

Claude Code, Cursor, VS Code and Codex: a token

On the Agent access card, give the token a name (for example "Claude Code, my laptop"), choose its scope, and press Create token.
Copy the token straight away. It is stored as a hash, so this is the only time it is shown.
Pick your client on the card. The setup snippet is filled in with your URL and token.

The snippets look like this, with your own values in place of the placeholders.

Claude Code, one command:

claude mcp add --transport http pangoling https://yourbrand.pangol.ing/api/mcp --header "Authorization: Bearer <your token>"

Claude Code can also sign in without a token. Run the same command without --header, then in Claude Code open /mcp, pick pangoling and choose Authenticate.

Cursor, in ~/.cursor/mcp.json:

{
  "mcpServers": {
    "pangoling": {
      "url": "https://yourbrand.pangol.ing/api/mcp",
      "headers": { "Authorization": "Bearer <your token>" }
    }
  }
}

VS Code, in .vscode/mcp.json:

{
  "servers": {
    "pangoling": {
      "type": "http",
      "url": "https://yourbrand.pangol.ing/api/mcp",
      "headers": { "Authorization": "Bearer <your token>" }
    }
  }
}

Codex, in ~/.codex/config.toml, with the token in your shell:

export PANGOLING_TOKEN=<your token>

[mcp_servers.pangoling]
url = "https://yourbrand.pangol.ing/api/mcp"
bearer_token_env_var = "PANGOLING_TOKEN"

Claude Desktop, through mcp-remote:

export PANGOLING_TOKEN=<your token>

npx mcp-remote https://yourbrand.pangol.ing/api/mcp --header "Authorization:Bearer ${PANGOLING_TOKEN}"

Not sure which file goes where? Use Copy setup prompt on the card and paste it to the assistant. It asks the assistant to add the server itself and confirm it is pointed at the right workspace.

Token scopes

Scope	What it allows
Read only	Reads everything, spends nothing.
Read and write	Also tracks keywords, saves drafts and brain notes, and runs paid lookups you ask for.

Neither scope can publish. Anything an assistant writes lands in Create as a draft, for a person to read and approve.

To stop a token, press Revoke next to it. Any assistant using it stops working immediately.

What you can ask for

The server ships ready-made prompts. Most assistants list them in their prompt or slash-command menu.

Prompt	What it does
site-checkup	A quick read of where the site stands: search overview, open fixes and site-audit findings.
keyword-research	Researches a topic, or your best opportunities overall, against your keywords, realistic wins and rival gaps.
competitor-analysis	Where a watched rival is ahead, where there is a real gap, and how their authority compares.
write-a-draft	Builds the brief for a keyword and writes the piece, saved as a draft for review.
review-queue	Walks the pieces waiting for sign-off and records your decisions.
seo-coach	A running conversation about your SEO, grounded in your own data.

You do not have to use them. Asking in plain words works too: "which pages are close to page one?" or "what did our main rival publish this month?"

Credits

Assistants spend from the same monthly credit pool as everything else, at the same prices. See Plans and credits. Each token also has a daily ceiling, so a looping agent cannot empty the pool in one afternoon. The card shows each token's calls and today's spend.

Plain HTTP API

If your agent does not speak MCP, the same tools are available over plain HTTP. The base URL is under Plain HTTP API on the Agent access card. Send the token as Authorization: Bearer <token>. A GET on the base URL lists every tool it can call.

Use pangoling from your AI assistant