Skip to main content
The server exposes 23 tools covering apps, surveys, analytics, responses, audience, and higher-level workflows. IDs accept either the short URL slug (e.g. a1b2c3d4) or the full UUID — list_apps and list_surveys are the easiest way to discover them.

Apps

Surveys

Insito uses a draft → live model. create_survey and update_survey write to the draft; nothing is shown to users until you run activate_survey. This mirrors the dashboard’s builder, so an agent and a teammate can’t step on each other’s live surveys.

Confirmations

activate_survey, update_brand_kit and delete_survey only run after you confirm. The server asks your MCP client to show a confirmation prompt (MCP elicitation), so the AI model can’t confirm on your behalf, even if text in a survey answer tells it to. This works the same on the local and the hosted server. If your client doesn’t support confirmation prompts, these three tools refuse and link to the dashboard page where you can do the same thing; all other tools work as usual. Confirmation prompts rely on your MCP client showing them honestly. As a second layer, deleting surveys also needs a key or OAuth grant with the opt-in Delete surveys permission.

Untrusted data

Survey answers, user IDs, user properties, and event and screen names come from your app’s users, or from anyone holding your public SDK key. Tools return them inside <untrusted_respondent_data> blocks with a note telling the model to treat them as data, never as instructions. Every tool also declares MCP annotations (readOnlyHint, destructiveHint, idempotentHint), so your client can ask before running anything that writes.

Question types

create_survey and update_survey take a questions array. Each question has a type and type-specific fields:

Analytics & responses

Platform and locale audience breakdowns aren’t available through the API yet — get_audience returns the user count and custom user-property distribution.

Composite tools

Higher-level tools that combine several API calls into one step:

Example prompts

Once configured, drive everything in natural language:
  • Discover — “List my Insito apps and their surveys.”
  • Events — “What events can I trigger a survey on in the Acme app?”
  • Create — “In the Acme app, create an NPS survey called ‘Onboarding NPS’ that fires on the onboarding_completed event, with a 0–10 NPS question and an open-text ‘What could be better?’ follow-up. Then make it live.”
  • Edit — “Add an end screen saying ‘Thanks!’ to the Onboarding NPS survey and publish it.”
  • Analyse — “What’s the NPS score for the Onboarding NPS survey, and how has it trended over the last 30 days?”
  • Browse — “Show me the 10 most recent responses for the Acme app on iOS.”
  • Segment — “What user properties do we capture in the Acme app, and how many users have each?”
  • Summarise — “Summarise the open-text feedback from the Onboarding NPS survey into themes.”
  • Compare — “Compare the Onboarding NPS survey’s score in June vs May.”
  • Design — “Suggest a survey to understand why users cancel, then create it in the Acme app.”

Errors

Tools return readable errors an agent can recover from — a duplicate trigger key, a validation problem, a missing ID, or an authentication failure all come back as plain text explaining what to fix.