Usercall

What Usercall does

The Usercall MCP server lets an agent arrange interviews with real users through Usercall. A study defines the research goal, optional business context, target number of interviews, languages, duration, interview mode, and optional visual material such as an image or Figma prototype. The study returns an interview link that can be shared with participants through external channels or product prompts.

Interview modes include voice, text, and voice-and-text. Results can include themes, summaries, insights, risks, and verbatim participant quotes. The server is suited to questions such as why users abandon onboarding, churn, misunderstand a feature, or fail a particular action.

How it works

The Usercall MCP server can run as a hosted MCP connection or as a local stdio process. The hosted service is available at https://mcp.usercall.co and uses OAuth sign-in. The local package is run with npx -y @usercall/mcp and authenticates with a Usercall API key.

For targeted research, an agent can inspect the product events Usercall has received, examine the schema for one event, and create a paused Research Trigger tied to an existing study. After a human activates the trigger in Usercall, the SDK invites matching users after the configured product event. Trigger filters use exact matches on event properties or user traits, with optional sampling and daily invite limits.

The trigger workflow depends on the Usercall SDK being installed in the product. If no events are available, get_trigger_sdk_setup provides snippets for installing the SDK, identifying users, and updating an allowlist. An analytics system such as PostHog or Mixpanel can identify behavior, but the supplied tools handle the Usercall study and trigger steps.

Setup and configuration

For local use, create an API key in Usercall under Home → Developer → Create API key, then provide it as USERCALL_API_KEY. Configure the MCP client to start npx -y @usercall/mcp over stdio. The README includes configurations for Claude Desktop and Cursor. Hosted connections should use the remote endpoint instead of the local JSON configuration.

A study requires key_research_goal. Optional fields include business context, target interviews, languages, duration, interview mode, voice gender, metadata, link variables, and study media. The research goal cannot be changed after creation. Creating a study does not itself run interviews or invite a participant; it returns an interview link for distribution.

Tools and capabilities

The Usercall MCP server exposes capabilities for:

• Creating and listing interview studies and retrieving study results.
• Listing observed product events from the previous 30 days.
• Inspecting event fields, types, and sample values.
• Checking supported trigger conditions and obtaining SDK setup snippets.
• Creating, listing, inspecting, updating, and deleting Research Triggers.
• Returning trigger status, activation URLs, invite counts, and interview counts.

Trigger conditions support one event, an exact property or trait match, a URL rule, and page dwell. Counts, sequences, absence conditions, time windows, and not-equals filters are not supported. Only event names already observed by Usercall can be used, and event matching is case- and type-sensitive. New or edited triggers are paused for activation; setting an active status directly is rejected, and editing an active trigger pauses it. Deleting a trigger is permanent, although completed interviews remain on the study.