Get started

Quickstart

Make your first authenticated REST request and connect your first MCP client.

Verified against the implementation ·

1. Create an API key

  1. 1

    Open the developer dashboard

    Sign in, open your PrompTessor workspace, and select API & MCP.

  2. 2

    Create a key

    Give the key a recognizable name, choose only the scopes the integration needs, and optionally set an expiration date.

  3. 3

    Copy it once

    Keys start with pt_live_. The secret is shown only at creation, so store it in a password manager or secrets service.

2. Inspect account capabilities

curl https://api.promptessor.com/v1/capabilities \
  -H "Authorization: Bearer pt_live_YOUR_API_KEY"

3. Generate a prompt

Every credit-consuming AI request requires a unique Idempotency-Key. Reusing the same key with the same body safely replays the stored response; reusing it with a different body returns an idempotency conflict.

curl -X POST https://api.promptessor.com/v1/prompts/generate \
  -H "Authorization: Bearer pt_live_YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: prompt-quickstart-001" \
  -d '{
    "message": "Create a launch plan for a privacy-first budgeting app",
    "mode": "planning",
    "outputMode": "ready_to_use",
    "targetModel": "ChatGPT",
    "language": "en"
  }'

Successful responses

A synchronous AI response includes the operation id, object, succeeded status, workflow data, requestId, usage charge, and—when the workflow stores history—a history link.

4. Connect an MCP client

Use https://mcp.promptessor.com/mcp as a Streamable HTTP server. Remote MCP requires OAuth; REST API keys are intentionally rejected on the MCP channel.

  1. 1

    Add a remote server

    Choose Streamable HTTP or HTTP in your client and enter the PrompTessor MCP URL.

  2. 2

    Authenticate

    Allow the client to discover PrompTessor OAuth, sign in, review the requested scopes, and approve access.

  3. 3

    Verify

    List tools or ask the client to use get_usage. Then try generate_prompt with a stable clientRequestId of at least eight characters.