Connect Custom GPT

1

Get Your API Key

Sign in to the Haiven dashboard to generate your unique API key. This connects your Custom GPT to your memory database.

Go to Dashboard →

Where to find it:

Dashboard → Settings → API Keys → "Generate New Key"

💡 Tip: Copy your API key somewhere safe. You'll need it in the next step.

2

Create Custom GPT

Go to ChatGPT and create a new Custom GPT with memory integration. You'll need ChatGPT Plus or Team.

Open GPT Builder →

⚠️ Requires: ChatGPT Plus ($20/mo) or Team plan. Free accounts cannot create Custom GPTs.

3

Add Memory Actions

In the GPT Builder, go to "Configure" → "Actions" → "Create new action" and paste this schema:

{
  "openapi": "3.1.0",
  "info": {
    "title": "Haiven Memory API",
    "version": "1.0.0"
  },
  "servers": [
    {
      "url": "https://api.safehaiven.com"
    }
  ],
  "paths": {
    "/api/memory/query": {
      "post": {
        "operationId": "queryMemories",
        "summary": "Search user memories",
        "requestBody": {
          "required": true,
          "content": {
            "application/json": {
              "schema": {
                "type": "object",
                "properties": {
                  "query": {
                    "type": "string"
                  },
                  "limit": {
                    "type": "integer",
                    "default": 5
                  }
                }
              }
            }
          }
        },
        "responses": {
          "200": {
            "description": "Relevant memories"
          }
        }
      }
    }
  },
  "components": {
    "securitySchemes": {
      "ApiKeyAuth": {
        "type": "apiKey",
        "in": "header",
        "name": "X-API-Key"
      }
    }
  },
  "security": [
    {
      "ApiKeyAuth": []
    }
  ]
}

🔑 Next: When prompted for authentication, select "API Key" and paste your key from Step 1.

4

Add System Instructions

In "Instructions", tell the GPT how to use memories. Copy and paste this:

You are a ChatGPT with perfect memory, powered by Haiven.

Before answering any user question, ALWAYS:
1. Call queryMemories with the user's question
2. Review returned memories for relevant context
3. Integrate that context naturally into your response

Never mention that you're querying memories - just use them seamlessly.

Examples of what to remember:
- User's role, company, projects
- Their preferences and working style
- Past decisions and reasoning
- Technical stack and tools they use
- Goals and challenges they've shared

Always prioritize recent memories over old ones when there's a conflict.

5

Test & Publish

Test your Custom GPT by asking it something personal. If it recalls context from previous conversations, you're done! Click "Save" and choose "Only me" for privacy.

Test conversation:

You: "What's my role?"

GPT: "You're a product manager at [YourCompany], working on..."