Configuring your agents

Mantle comes with a set of pre-built system agents designed for common tasks—support, content writing, customer insights, and more. These agents are already configured with sensible defaults, so you can start using them right away. This guide shows you how to configure them for your specific needs.

System agents

When you open Agents in your Mantle sidebar, you'll see the system agents available to your organization. These are ready to use out of the box:

• Support Sidekick — Analyzes support tickets, researches your docs and codebase, and drafts responses for your team to review. Follows a structured workflow: gather context, research the issue, classify it, then draft a response or recommend escalation. See AI-Powered Support for details on how it works in the help desk.

• Content Writer — Drafts documentation, help articles, and marketing content based on your product and existing docs.

• Customer Insights — Answers questions about your customer data, metrics, and business health. Good for ad-hoc analysis.

Each system agent comes with a default mandate, recommended tools, and connections already set up. You can use them as-is or customize them to better fit your workflow.

Configuring an agent

Click on any agent to open its configuration page. Here you can customize how it works.

Mandate

The mandate is the agent's job description—it defines what the agent should do and how it should behave. System agents come with default mandates that work well for most cases, but you can customize them to:

• Match your company's tone and voice

• Add specific guidelines for your product

• Include examples of good responses

• Set boundaries on what the agent should and shouldn't do

You can also click Generate to have AI help you write or refine the mandate. Describe what you want the agent to do, and Mantle generates a detailed mandate based on your description and any existing agents in your organization.

Tools

Tools give your agent capabilities—access to customer data, ability to run reports, create content, etc. Click Manage tools to see what's available.

Tools are organized by category:

• Core — Customer data, subscriptions, plans, transactions, app events, deals, tasks

• Metrics — Analytics and reporting (MRR, churn, growth, retention, traffic sources, funnel, reviews)

• Help Desk — Search tickets, view ticket details, create internal notes, search saved replies

• Flows — View and manage automation flows

• Affiliates — Affiliate program management and payouts

• Docs — Documentation management, page creation, publishing

In addition to category tools, agents have access to special Mantle tools:

• Search docs — Search your published documentation

• Create insight — Surface insights in the Mantle UI

• Create ticket response suggestion — Draft a suggested reply for a support ticket

• Create saved reply suggestion — Suggest new saved replies based on ticket patterns

• Create docs suggestions — Generate documentation improvement suggestions

• Email notify — Send email notifications to team members

• Slack notify — Send notifications to Slack

• Call sub-agent — Invoke another agent to handle a subtask

System agents come with relevant tools already enabled. You can add more if your agent needs additional capabilities, or remove tools to keep it focused.

Rules

Rules provide additional context beyond the mandate. They're uploaded as files that the agent can reference when working. Use them for:

• Detailed documentation the agent should reference

• Company-specific guidelines or terminology

• Edge cases and how to handle them

• Product-specific knowledge

• Response formatting and tone requirements

Click Add rule to create one. Each rule has a name and description that helps the agent decide when to reference it.

When the agent runs, it intelligently selects which rules are relevant to the current task. If you have only a few rules, it loads them all. With more rules, it uses the name and description to pick the most applicable ones and keeps the rest available on demand.

Codebase

For technical products, connecting your codebase lets the agent understand your actual implementation. When customers ask technical questions or report bugs, the agent can search your code to provide accurate answers.

Click Add codebase and choose:

• GitHub — Connect directly from GitHub

• GitLab — Connect from GitLab

• Upload — Upload a ZIP file of your code

Once connected, Mantle indexes your code for search. The agent gets access to multiple search tools:

• Semantic search — Find code by meaning (e.g., "where is payment processing handled?")

• Text search — Find specific function names, error messages, or patterns

• Hybrid search — Combine both approaches

• File browsing — List files, view structure, find references and definitions

Mantle supports a wide range of languages including JavaScript, TypeScript, Python, Java, Go, Rust, Ruby, PHP, C#, Swift, and many more.

External MCPs

If you want your agent to access data beyond Mantle, you can connect external Model Context Protocols (MCPs). These let your agent access:

• Your own app's API

• Third-party tools (CRMs, helpdesks, etc.)

• Internal systems and databases

Click Add MCP and provide the server URL and authentication details. Credentials are stored with field-level encryption.

Visibility

Control who on your team can see and use this agent. In the Details sidebar, find the Visibility section:

• Organization — Anyone in your organization can use this agent

• By role — Only team members with specific roles (Owner, Admin, Member)

• By feature access — Only users with access to specific Mantle features

• Private — Only you can see and use this agent

Agent jobs

Agent jobs let you schedule your agents to run automatically or trigger them on specific events—without setting up a full Flow.

Each job has:

• Triggers — When the job runs. Options include scheduled (daily/weekly at a specific time and timezone) or event-based (e.g., when a customer uninstalls, a ticket is created, etc.)

• Instructions — What the agent should do when the job runs. This is separate from the mandate—it's the specific task for this job.

• Actions — What happens after the agent finishes. Options include creating insights, generating ticket response suggestions, sending email notifications, or posting to Slack.

For example, you could create a job that:

• Runs every Monday morning at 7 AM

• Tells the agent to "Analyze this week's support tickets and identify the top 3 recurring issues"

• Sends the result as an email to your support lead

Or an event-triggered job that:

• Fires when a customer uninstalls your app

• Tells the agent to "Analyze this customer's recent tickets and usage to identify why they may have left"

• Posts the findings to a Slack channel

Testing and debugging

After configuring your agent, use the testing tools in the top-right of the agent configuration page to make sure it works as expected.

Chat with agent

Click Chat with agent to open Mantle's chat interface with your agent selected. This is the fastest way to test conversational interactions—just type questions and see how the agent responds.

Try asking questions like:

• "How many customers do I have?"

• "What's my current MRR?"

• "Show me my top 5 customers by revenue"

For agent-specific testing:

• For Support Sidekick: "Summarize the open tickets from today"

• For Content Writer: "Draft a help article about getting started"

• For Customer Insights: "Which customers are at risk of churning?"

If your agent can't answer something, check which tools are enabled.

Agent debugger

Click Agent debugger to test your agent with more control. The debugger lets you simulate how the agent will run in flows or automated contexts.

In the debugger, you can:

• Set resource context — Choose whether to run the agent with no context, or with a specific customer or ticket selected. This simulates how flows pass data to agents.

• Write instructions — Tell the agent exactly what you want it to do, just like the instructions field in a flow action.

• Define output schema — Optionally specify fields the agent should return. This tests structured output for flow integrations.

View usage

Click View usage to see your agent's activity history. This shows every time the agent ran, including:

• Run status (completed, error, running)

• When each run started and how long it took

• The instructions that triggered each run

• Full run details including tool calls and responses

Use the activity view to identify issues—if runs are failing, you can see exactly what went wrong and adjust your agent's configuration.

Creating a custom agent

While system agents cover most common use cases, you can create custom agents for specialized needs. Click Add agent from the Agents page.

You'll need to provide:

• Name — Something descriptive like "Onboarding Specialist" or "Churn Analyzer"

• Handle — A short identifier for chat shortcuts (e.g., /onboard)

• Apps — Choose whether the agent can access all apps or specific ones

• Mandate — Define the agent's purpose and behavior

After creating, configure tools, rules, and connections as described above.

Using your agents

Once configured, you can use agents in multiple ways:

• Mantle chat — Use the handle shortcut (e.g., /support) to switch to an agent in the chat interface

• Flows — Add agents as actions in automation flows to analyze data and make decisions

• Agent jobs — Schedule agents to run on a timer or in response to events

• Help desk — The Support Sidekick runs directly from the ticket sidebar

• External tools — Connect via MCP to Cursor, Claude Desktop, or other AI tools

Learn more: Using agents in Flows

Tips for getting the most out of agents

• Start with system agents — They're already tuned for common tasks. Customize them rather than starting from scratch.

• Customize the mandate — Add your company voice, product context, and specific guidelines. The more detail you provide, the better the agent performs.

• Use Generate to get started — If writing a mandate from scratch feels daunting, describe what you want and let AI draft it for you.

• Connect your codebase — For technical products, this dramatically improves accuracy for bug investigations and how-to questions.

• Add rules for edge cases — If the agent mishandles something, add a rule to clarify. Rules are easier to iterate on than rewriting the entire mandate.

• Use the debugger before going live — Test with real ticket and customer data in the debugger before enabling automated responses.

• Check activity regularly — Review the usage logs to see how the agent performs and catch issues early.

• Iterate over time — Refine mandates and rules based on what works. Good agents are built through iteration, not in one shot.

• AI-Powered Support — How the Support Sidekick works in the help desk

• Agent use cases — Real examples of agents in action

• What's in an agent — Understanding mandates, tools, rules, and more