Creating Your First Agent

Prerequisites

Before creating your first agent, ensure you have:

Admin access to your Bloom organization
API credentials for the services you want to integrate
Basic understanding of HTTP APIs

Step 1: Create an Integration

Navigate to Integrations

Go to the Integrations tab in your Bloom dashboard

Add Service

Click “Add Integration” and select your desired service (e.g., OpenAI, GitHub, Serper)

Configure Credentials

Enter your API key and any required configuration for the service

Store your API keys securely. Bloom encrypts them but ensure they have appropriate permissions.

Test Connection

Use the test button to verify your integration is working correctly

Step 2: Define Scopes

Scopes control what your agent can access. Create granular permissions:

Go to Scopes Tab

Navigate to the Scopes section in your dashboard

Create New Scope

Click “Create Scope” and give it a descriptive name

Configure Permissions

Set up the allowed HTTP methods and API paths:

{
  "service": "openai",
  "methods": ["POST"],
  "paths": ["/chat/completions", "/embeddings"]
}

Save Scope

Review and save your scope configuration

Step 3: Create Your Agent

Navigate to Agents

Go to the Agents tab in your dashboard

Add New Agent

Click “Create Agent” and provide:

Name: Descriptive name for your agent
Description: What this agent does

Assign Scopes

Select the scopes you created in Step 2

You can assign multiple scopes to give your agent access to different services

Generate Agent ID

Copy your agent ID - you’ll need this for implementation

Step 4: Implement in Code

Choose your integration method:

Direct API
MCP Server

import requests
import os

AGENT_ID = os.getenv("BLOOM_AGENT_ID")
ORG_API_KEY = os.getenv("BLOOM_ORG_API_KEY")

def call_openai(prompt):
    response = requests.post(
        "https://iam.bloomtechnologies.app/https://api.openai.com/v1/chat/completions",
        headers={"Authorization": f"Bearer {ORG_API_KEY}"},
        json={
            "agent_id": AGENT_ID,
            "model": "gpt-3.5-turbo",
            "messages": [{"role": "user", "content": prompt}],
            "max_tokens": 150
        }
    )
    return response.json()

# Test your agent
result = call_openai("Hello from my Bloom agent!")
print(result)

Add to your mcp_config.json:

{
  "mcpServers": {
    "openai": {
      "command": "npx",
      "args": ["-y", "bloom-mcp-wrapper", "@modelcontextprotocol/server-openai"],
      "env": {
        "BLOOM_AUTH": "bloom_{ORG_KEY}_agent_{AGENT_ID}"
      }
    }
  }
}

Verification

Test that everything is working:

Check Agent Status: Verify your agent appears in the dashboard
Test API Call: Make a simple request through the proxy
Review Logs: Check the activity logs in your dashboard
Validate Scopes: Ensure your agent can only access permitted endpoints

If you encounter authentication errors, double-check your environment variables and scope assignments.

Common Issues

403 Forbidden Error

Your agent likely doesn’t have the required scopes. Check:

Agent has the correct scopes assigned
Scope includes the HTTP method you’re using
Scope includes the exact API path

Invalid Agent ID

Verify:

BLOOM_AGENT_ID environment variable is set correctly
Agent ID exists in your dashboard
Agent hasn’t been deleted or deactivated

Integration Not Found

Ensure:

Integration exists and is properly configured
Service name in URL matches integration name
API credentials are valid and have necessary permissions

Get Started

Guides

Integration Methods

Security

Prerequisites

Step 1: Create an Integration

Step 2: Define Scopes

Step 3: Create Your Agent

Step 4: Implement in Code

Verification

Common Issues

Next Steps

​Prerequisites

​Step 1: Create an Integration

​Step 2: Define Scopes

​Step 3: Create Your Agent

​Step 4: Implement in Code

​Verification

​Common Issues

​Next Steps

Prerequisites

Step 1: Create an Integration

Step 2: Define Scopes

Step 3: Create Your Agent

Step 4: Implement in Code

Verification

Common Issues

Next Steps