Connectors

Connect AI agents to external services like Jira, Salesforce, Zendesk, and more

Connectors extend your AI agents with access to external services and data sources. Each connector is an MCP server that provides tools for interacting with a specific platform.

Available Connectors

Jira & Confluence

Manage issues, sprints, boards, and Confluence pages with 51 tools for complete Atlassian integration

Salesforce CRM

Query and modify Salesforce data with SOQL, create records, and access object metadata

Zendesk Support

Manage support tickets, add comments, and access Help Center knowledge base

Bright Data Search

Search the web, scrape content, and automate browsers with anti-blocking technology

Exa AI Search

Neural search engine with semantic understanding for research and content discovery

Playwright Browser

Automate web browsers for testing, screenshots, and complex web interactions

Integration Methods

There are three ways to connect external services to your Heroku AI agents:

Method 1: Deploy MCP Server to Heroku (Recommended)

Deploy an MCP server as a Heroku app and register it with your AI model. Tools become available through the /v1/agents/heroku endpoint.

# Deploy the MCP server
heroku create my-jira-mcp
git push heroku main

# Attach your AI model
heroku addons:attach my-app::INFERENCE -a my-jira-mcp

# Tools are now available via /v1/agents/heroku

Best for: Production deployments, full control over configuration, no external dependencies.

Method 2: Connect to Remote MCP Servers

Some services offer hosted MCP servers (e.g., Atlassian’s mcp.atlassian.com). Connect using the Vercel AI SDK or Pydantic AI as an MCP client.

Vercel AI SDK

import { experimental_createMCPClient } from "ai/mcp";
import { createHeroku } from "@heroku/ai-provider";

const mcpClient = await experimental_createMCPClient({
  transport: {
    type: "sse",
    url: "https://mcp.atlassian.com/v1/sse",
  },
});

const tools = await mcpClient.tools();
const heroku = createHeroku();

const response = await generateText({
  model: heroku("claude-4-sonnet"),
  tools,
  prompt: "List my open Jira issues",
});

Pydantic AI

from pydantic_ai import Agent
from pydantic_ai.mcp import MCPServerSSE

mcp_server = MCPServerSSE("https://mcp.atlassian.com/v1/sse")

agent = Agent(
    "heroku:claude-4-sonnet",
    mcp_servers=[mcp_server]
)

async with agent.run_mcp_servers():
    result = await agent.run("List my open Jira issues")
    print(result.output)

Best for: Using official hosted MCP servers, quick prototyping, when you don’t need custom configuration.

Method 3: Direct API Integration

For services without MCP support, build custom tools that call APIs directly. See Custom Tools for details.

Connector Comparison

Connector	Official Remote MCP	Deploy to Heroku	Tools
Jira & Confluence	mcp.atlassian.com	Community server	51 tools
Salesforce	Beta (GA Feb 2026)	Community server	5 tools
Zendesk	Not available	Community server	6 tools
Bright Data	Not available	Official server	10+ tools
Exa AI	Not available	Community server	6 tools
Playwright	Not available	Official server	8+ tools

Quick Start

1. Choose a Connector

Browse the available connectors above and click through to the detailed documentation.

2. Deploy or Connect

Each connector page includes:

• One-click “Deploy to Heroku” button (for Heroku deployment)
• Configuration for remote MCP connection (where available)
• Required environment variables and API keys

3. Use with Agents

Once connected, include the MCP server in your agent requests:

response = client.chat.completions.create(
    model=os.getenv("INFERENCE_MODEL_ID"),
    messages=[
        {"role": "user", "content": "Create a Jira ticket for the login bug"}
    ],
    extra_body={
        "heroku": {
            "mcp_servers": ["jira-mcp"]
        }
    }
)

Building Custom Connectors

Need to connect to a service not listed here? You can build your own MCP server:

1. Implement the MCP protocol
2. Define tools with clear descriptions and input schemas
3. Deploy to Heroku with an mcp: process in your Procfile
4. Attach your AI model to register the tools

See Custom Tools for a complete guide.

Related Resources

MCP Overview

Learn about the Model Context Protocol

Agents API

Full API reference for using tools with agents