API Keys (MCP) setup guide

What it does in ContentPulse: Lets external AI agents — Claude Desktop, ChatGPT (via custom GPT), Cursor, or any tool that speaks the Model Context Protocol (MCP) — read and act on your ContentPulse workspace. Generate pieces from a chat, list pending QA items, publish on a voice command, etc. Status: Available Tier: Pro plan allows up to 3 concurrent connections; Agency and Enterprise are unlimited.

Before you start (prerequisites)

Account requirement: A ContentPulse Pro plan or higher.
Client: An MCP-capable client — most commonly Claude Desktop (https://claude.ai/download) or a custom GPT in ChatGPT. The MCP protocol is also supported by Cursor, Zed, and Continue.
Cost: Free on ContentPulse's side. Third-party costs depend on the client (Claude Desktop is free with a Claude account; ChatGPT custom GPTs require ChatGPT Plus).

Step-by-step setup

1. In ContentPulse — issue a key

Go to https://contentpulse.helloaurora.ai/settings/api-keys.
Click New Key.
Give it a descriptive name — what device or agent will use this key. Examples: "Claude desktop — laptop", "ChatGPT — campaign GPT", "Cursor — dev machine". The name only affects how you identify the key in the dashboard; it has no functional effect. [Screenshot to be added](screenshot:API keys page with New Key form)
Click Issue Key. ContentPulse shows you a string like rf_sk_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx exactly once — save it now, this is the only time it will be shown. We never store the raw key; if you lose it, you have to revoke and re-issue.
Copy the key (button on the right of the secret banner). Click I've saved it to dismiss the banner.

2. Configure your MCP client

Claude Desktop:

Open Claude Desktop → Settings → Developer → Edit Config.
This opens claude_desktop_config.json in your default editor. Add a contentpulse entry under mcpServers:

{
  "mcpServers": {
    "contentpulse": {
      "url": "https://contentpulse.helloaurora.ai/api/mcp/sse",
      "headers": {
        "Authorization": "Bearer rf_sk_your_key_here"
      }
    }
  }
}

Save the file and fully quit Claude Desktop (Cmd+Q on Mac, exit tray icon on Windows).
Reopen Claude Desktop. You should see a hammer/tools icon in the input bar — click it and you'll see ContentPulse tools listed (list_pieces, publish_piece, generate_from_source, etc.). [Screenshot to be added](screenshot:Claude Desktop with ContentPulse tools dropdown visible)

ChatGPT (custom GPT with actions):

Go to https://chat.openai.com/gpts/editor and create a new GPT.
In the Configure tab, scroll to Actions → Create new action.
Use ContentPulse's OpenAPI spec at https://contentpulse.helloaurora.ai/api/mcp/openapi.json — paste the URL into the Import from URL field.
In Authentication, choose API Key, Bearer, and paste your rf_sk_... key.
Save the GPT. Test by asking it "list my pending pieces in ContentPulse".

Other MCP clients: Use the same SSE endpoint (https://contentpulse.helloaurora.ai/api/mcp/sse) and Bearer auth. See your client's MCP server configuration docs.

3. Verify it's working

In Claude Desktop, type "Using ContentPulse, list my last 5 published pieces." Claude should call list_pieces and return results.
On the ContentPulse Settings → API Keys page, the key's "Last used" timestamp updates within a few seconds of the first request.

What you can do with it once connected

The MCP surface exposes these capabilities to your agent:

Read: list sources, list pieces, get piece details, check QA status, list scheduled queue items.
Generate: create new pieces from a source URL or pasted text in your trained voice.
Edit: modify a piece's title or body, re-run QA.
Publish: publish a piece to a connected destination immediately or schedule for later.
Inspect: check usage limits, see plan tier, list integrations.

Examples of natural-language prompts that work once connected:

"Pull yesterday's podcast from my Riverside feed into ContentPulse and generate 3 LinkedIn posts in my voice."
"Which pieces in QA failed last week and why?"
"Reschedule everything in the Wednesday queue to next Monday."

Common pitfalls

Pitfall: "401 Unauthorized" from Claude Desktop. Fix: Three likely causes — (1) you copied the key with an extra space at the start or end (re-copy from your saved password manager), (2) the key was revoked on the API Keys page, or (3) the Bearer prefix is missing — in claude_desktop_config.json the value must be exactly "Bearer rf_sk_..." with one space.
Pitfall: Claude says "tools not available" after editing config. Fix: Claude Desktop must be fully quit and reopened, not just the window closed. On Mac: Cmd+Q. On Windows: right-click the tray icon → Quit.
Pitfall: Agent calls a tool but gets "plan tier insufficient". Fix: Some tools (publishing, bulk operations) are tier-gated. Check your plan at /settings/billing.
Pitfall: I lost the key. Fix: Keys are non-recoverable by design. Revoke the lost one on the API Keys page and issue a new one.
Pitfall: Pro plan says "up to 3 connections" — what counts as a connection? Fix: Each API key is one connection. Three keys = three connections. The limit is on key count, not on simultaneous requests.

Restrictions

The MCP surface enforces the same row-level security as the web UI — your agent can only see and act on your workspace.
Some destructive actions (delete a piece, disconnect a platform) require human confirmation; the agent surfaces a prompt rather than acting silently.
Pro plan: max 3 active keys. Agency+: unlimited.
Rate limit: 100 MCP requests per minute per key.

Need help?

Email [support@helloaurora.ai](mailto:support@helloaurora.ai) — we'll walk you through it. We can also send a working claude_desktop_config.json template if you let us know your OS and Claude Desktop version.