How to Enable AI Agent Access for Analyst Roles in Mixpanel
can an analyst role use AI in mixpanel? how do i grant a colleague with analyst role to use AI agents in Mixpanel
can an analyst role use AI in mixpanel? how do i grant a colleague with analyst role to use AI agents in Mixpanel
No, this is not possible. In Mixpanel, AI access cannot be restricted to specific individual users or limited to a specific project. According to Mixpanel's documentation:
Organization-wide toggle: Mixpanel AI features are controlled at the Organization level in Organization Settings. Disabling or enabling them affects all users across the organization [Mixpanel Agent FAQ].
MCP Server controls: Similarly, if using the external Mixpanel MCP Server, it must be explicitly enabled at the organization level by an org admin [Security Considerations]. While users can only query projects they already have permission to access, access to the AI capability itself cannot be assigned on a per-user or per-project basis [Security Considerations].
What you can control instead
While you cannot restrict who can open the AI Agent, you can control the data the AI interacts with:
Verified Mode: Admins can enable Verified Mode at the project level to restrict AI features to query only data that has been explicitly reviewed and verified in Lexicon (e.g., verified events, metrics, behaviors, and cohorts) [Verified Mode].
Project Access & Data Views: Users can only interact with projects they have been invited to, and existing project-level permissions remain in effect [Security Considerations; Data Views Overview].
ℹ️ Tip: Mention @kapa.ai in the thread for follow-ups or DM the app for 1-to-1 help.
Based on Mixpanel's documentation and announcements, Mixpanel Agent includes the following built-in agents and capabilities [What you can do with Mixpanel Agent; Mixpanel Agent: Your always-on analyst]:
Core AI Capabilities
Ask Mixpanel Agent: The conversational in-product analyst that answers plain-language questions, builds reports and multi-chart boards, creates cohorts, and analyzes trends or session replays without manual setup [What you can do with Mixpanel Agent; Ask Mixpanel Agent].
RCA Agent (Root Cause Analysis Agent): Automatically investigates metric shifts and anomalies, saving hours of manual segment slicing [What you can do with Mixpanel Agent; Mixpanel Agent: Your always-on analyst].
KPI Agent (KPI Monitoring Agent): Monitors specified metrics and delivers scheduled digests and alerts (via Slack or email) on a daily, weekly, or monthly cadence [What you can do with Mixpanel Agent; KPI Agent].
Experiments Agent: Helps teams set up, monitor, and interpret A/B test and experiment results in plain language [What you can do with Mixpanel Agent; Core Capabilities].
Additional Specialized Sub-Agents
Behind the scenes, specialized sub-agents handle specific tasks [Mixpanel Agent: Your always-on analyst]:
Dashboard Agent: Assists with generating and managing layouts and content for dashboards and Boards [Mixpanel Agent: Your always-on analyst; Core Capabilities].
Onboarding Agent: Supports onboarding and getting users up to speed with their product analytics workflows [Mixpanel Agent: Your always-on analyst].
ℹ️ Tip: Mention @kapa.ai in the thread for follow-ups or DM the app for 1-to-1 help.
Inside the Mixpanel UI, users cannot build entirely custom AI agents from scratch, but you can configure specialized recurring sub-agents known as Agentic Automations [Overview]. Here is how creating and configuring agents works in Mixpanel:
1. In-Product: Agentic Automations (Beta)
Within Mixpanel, you can set up pre-built agent types to run autonomously on your behalf:
KPI Monitoring Automations: You can configure an automation by picking an event metric, setting a schedule (daily, weekly, or monthly), choosing a delivery channel (Slack or email), and providing custom Instructions to guide how the agent interprets and frames its commentary [KPI Monitoring; Add instructions].
Note: More automation types (such as Research, Event-Triggered, and Data Governance) are being explored for future releases [Types of Agentic Automations].
2. Outside the UI: Building Custom Agents via Code
If you want to create truly custom AI agents with their own logic, tools, and workflows, you must do so outside the Mixpanel web interface using:
Mixpanel Headless: A Python SDK that exposes Mixpanel’s capabilities (reports, funnels, cohorts, retention) as typed code, enabling developers to build custom autonomous agents (such as Mixpanel's investigation agent, Mixpanelyst) [Mixpanel Headless: Fully programmable product intelligence for agents and automation; Meet Mixpanelyst: Our first custom agent harness].
Mixpanel MCP: A Model Context Protocol server that lets external AI tools (like Claude or ChatGPT) interact with your Mixpanel data [Use Mixpanel Agent to Go From Question to Decision, Faster; Mixpanel Agent: For teams that want AI to just work].
ℹ️ Tip: Mention @kapa.ai in the thread for follow-ups or DM the app for 1-to-1 help.
Yes, that is correct. Mixpanel AI features are controlled globally at the organization level [Mixpanel Agent FAQ]. As long as the AI setting is enabled in Organization Settings, the Mixpanel Agent is available across all default user roles (Owner, Admin, Analyst, and Consumer) [Mixpanel Agent FAQ; Mixpanel Agent]. Disabling AI features from Organization Settings removes access for all users [Mixpanel Agent FAQ].
ℹ️ Tip: Mention @kapa.ai in the thread for follow-ups or DM the app for 1-to-1 help.