The MCP connector catalog now includes five additional apps: Clay, Lemlist, Airtable, Composio, and Reap. These connectors extend Managed Agents and Ask AI beyond native actions by giving them controlled access to tools available in connected third-party accounts.

This guide explains what each new connector is designed for, how MCP apps work with Managed Agents and Ask AI, how to select the tools an agent can use, and how to test a connector before using it in live automation.

Clay, Lemlist, Airtable, Composio, and Reap are now available in the MCP connector catalog. They join the existing and growing collection of third-party integrations available to supported AI experiences.

What are the New MCP Connectors?

MCP, or Model Context Protocol, provides a standardized way for AI agents to connect to external applications and use the tools those applications expose. MCP app connectors allow Managed Agents and Ask AI to work with authenticated third-party services without requiring you to manually build every integration.

Once an MCP app is connected, you control which available tools the AI can use. A Managed Agent can call those tools when its trigger and instructions require them, while Ask AI can use enabled connector tools while responding to your requests.

The latest catalog expansion adds Clay, Lemlist, Airtable, Composio, and Reap.

Key Benefits of MCP Connectors

MCP connectors extend AI automation by letting agents securely access selected capabilities from third-party applications. This makes it possible to build richer automations without creating a separate custom integration for every use case.

• More External Actions: Let AI agents work with supported tools in connected third-party apps.
• Controlled Tool Access: Choose which tools each connected app makes available to the agent.
• Secure Connections: Authenticate through the connector's supported OAuth or secure sign-in flow instead of placing credentials in agent instructions.
• Multi-App Agents: Connect multiple MCP apps to a Managed Agent when one workflow needs to work across several systems.
• Faster Automation: Add prebuilt app capabilities without manually configuring individual API calls for every operation.
• Broader AI Workflows: Combine account events, AI reasoning, and external applications in one agent workflow.

Five New MCP Connectors

Each connector expands the types of work an AI agent can perform. Choose the app that matches the external data, outreach, record management, multi-app automation, or content-production task you want the agent to handle.

How MCP Connectors Work with Managed Agents

A connector gives the agent access to external tools, but it does not decide when the agent should run or what the agent should do. Managed Agents combine triggers, instructions, apps, capabilities, and optional Knowledge Bases so each part of the configuration has a clear role.

Triggers Determine When the Agent Runs

Triggers start a Managed Agent when the configured event occurs. For example, an agent might run when a contact is created, an opportunity is created, or a lead-generation form is submitted.

Apps Define Which External Systems the Agent Can Use

The Apps area contains the MCP connectors assigned to the Managed Agent. A single agent can use more than one connector when its instructions require actions across different external applications.

Instructions Tell the Agent What to Do

Clear instructions define when an external tool should be used, what information should be passed to it, and what outcome the agent should produce. Giving an agent access to a connector does not mean every available tool will be called on every run.

Using MCP Connectors in Managed Agents vs. Ask AI

Managed Agents and Ask AI can both use MCP connectors, but they serve different automation patterns. Managed Agents are designed for triggered, autonomous work, while Ask AI uses connected tools in response to a user's request.

How To Setup and Use New MCP Connectors

Proper setup ensures the AI has access only to the external account and tools required for its task. Connect the app, review its permissions, select the appropriate tools, and test the behavior before relying on the connector in production.

Option 1: Connect an MCP App to a Managed Agent

Managed Agents use MCP apps as part of event-driven automation. The app should be attached to the specific agent that needs its tools so access remains aligned with the agent's purpose.

Step 1: Open Managed Agents

Managed Agents are configured from Agent Studio, where you can define the agent's instructions, triggers, apps, capabilities, Knowledge Bases, testing, and publishing state.

1. Go to AI Agents → Agent Studio.
2. Open Managed Agents.
3. Create a new Managed Agent or open an existing agent.

Step 2: Add an MCP App

The Apps area determines which external connectors are available to the agent. Only add connectors that are relevant to the automation the agent is expected to perform.

1. Locate the Apps area.
2. Click Add app.
3. Browse or search the MCP catalog.
4. Select Clay, Lemlist, Airtable, Composio, Reap, or another supported connector.

Step 3: Connect the External Account

Authentication allows the connector to act within the permissions granted by the external service. Use the supported OAuth or secure sign-in flow presented by the connector.

1. Select the MCP app.
2. Click the option to connect or add an account.
3. Sign in to the applicable third-party account.
4. Review and authorize the requested access.

Step 4: Choose the Tools the Agent Can Use

Tool selection controls the specific actions exposed to the agent. Limiting access to the tools required for the use case makes the agent's capabilities easier to understand, test, and govern.

1. Review the available tools for the connector.
2. Select the tools the agent needs.
3. Leave unrelated tools disabled when they are not required.
4. Click Save.

Step 5: Update the Agent Instructions

The agent needs clear guidance about when and why to use the connector. Good instructions reduce unnecessary tool calls and make the expected outcome easier to test.

Include details such as:

• Which situations require the external app.
• What data the agent should read or send.
• Which actions it is allowed to perform.
• When the agent should ask for clarification.
• What the agent should do when the connector returns an error or no result.

Step 6: Configure the Trigger

Triggers determine when a Managed Agent begins its work. Choose an event that provides the right context for the connected app action.

For example, a lead-enrichment agent could run when a contact is created, while a follow-up agent could run after a form submission or another supported event.

Step 7: Test the Agent

Testing confirms that the correct connector and tool are selected, the external account returns the expected information, and the agent follows the restrictions in its instructions.

1. Click Test Agent.
2. Use representative test data.
3. Confirm the correct MCP tool is invoked.
4. Verify the expected result appears in the external app.
5. Refine the instructions or tool permissions if the result is not correct.

Step 8: Publish the Managed Agent

Publishing makes the tested configuration available for live execution. Publish only after confirming the trigger, external account, tool permissions, and agent instructions behave as intended.

Option 2: Connect an MCP App in Ask AI

Ask AI connectors are designed for user-initiated tasks. Connect the app once, control which tools are enabled, and then ask Ask AI to perform supported work using that connected account.

1. Open Ask AI.
2. Go to Customize → Connectors.
3. Select Manage Connectors.
4. Find the MCP app you want to use.
5. Connect and authenticate the applicable third-party account.
6. Review and enable the tools Ask AI should be allowed to use.
7. Enable the connector.
8. Start a chat and give Ask AI a clear request that requires the connected app.

Security and Access Best Practices

MCP connectors can allow AI to take real actions in external systems, so connector access should match the agent's exact responsibility. Limiting accounts, tools, and instructions reduces unnecessary access and makes production behavior easier to audit.

• Grant only required access: Authorize the minimum third-party permissions necessary for the agent's job.
• Select only required tools: Avoid enabling every available tool when the agent needs only a small subset.
• Keep credentials out of prompts: Use the connector authentication experience rather than placing passwords, API keys, or login details in agent instructions.
• Write explicit guardrails: Tell the agent which actions it may perform and which situations require confirmation or escalation.
• Test with realistic inputs: Verify read and write actions before exposing the agent to live triggers.
• Review third-party requirements: Some apps may require their own account, plan, permissions, or feature access.
• Recheck tools after connector updates: Available tool lists can evolve as MCP integrations add or change capabilities.

Troubleshooting MCP Connectors

Most MCP connector issues come from authentication, third-party permissions, tool selection, or unclear agent instructions. Check each layer separately to identify whether the connection, tool, or agent behavior needs to be adjusted.

Frequently Asked Questions

Need Help?