External AI agents

Bring your own AI agent powered by Claude, Codex, or another AI provider into Front to integrate it into your team's workflows like any human teammate. AI agents work seamlessly with features such as analytics and rules, allowing you to configure auto-assignment workflows to maintain complete control over whether conversations are handled by a human or AI.

You must have company admin permissions to set up AI agents.

Please review the instructions in Front’s developer documentation for all information on how to set up and manage external agents in Front.

Front’s documentation assumes you have already created an AI agent outside of Front, using tools like Claude, ChatGPT, n8n, industry-specific AI solutions, or any custom-built solution. The agent is hosted external to Front and has the skills necessary to evaluate customer conversations, access the sources of information it needs to make decisions, and draft replies.

After creating your external AI agent in Front, see the main AI agents guide to review how to:

• Assign conversations to your AI agent

• Monitor and track AI agent analytics

• Optimize AI agent performance

• Find answers to frequently asked questions

External agent traffic through MCP does not consume the customer’s Core API allowance or destabilize existing API integrations. We have aggregate caps of 1,000 calls/hour per AI teammate and 10,000 calls/hour per workspace. For individual tools see below.

There are two primary advantages: Governance and Performance.

Governance:

• AI teammate identity that is deeply embedded across Front

• Granular permission system for what the AI agent can and can't do and where it takes action, enforced directly from Front

• Automatic and manual invocation of the agent anywhere in Front without awkward backlash rules, etc.

Performance

• AI-optimized MCP server connection so you don't have to burn tokens or processing steps parsing the API response

• Separate bandwidth lane for the MCP server so you don't have to burn through your API rate limit or put other integrations at risk

• Observability and repair loop

• Ability for human teammates to provide feedback to the agent

• Automatic performance tracking of agent based on draft drift and other quality markers and Front analysis

• Insights and analytics and dashboards displaying how the agent is performing

• Future plans to help you test and automatically send performance data to your agent for a full circle, low effort repair loop

For external agents, Front can't see what the agent knows or why it answered the way it did, so the ‘Sources’ indicator is not included in those messages. Front can only show what is passed via the MCP - which are just the actions the agent takes in the conversation.

See this FAQ section for common questions that apply to AI agents.

This feature is available on all plans.