For product managers
Draft one between two meetings
Describe the survey to the assistant you already have open, read the four questions it drafted, fix the one that is off. Live before the next call starts.
HomeSurveys
Your editor knows the product.
Let it write the questions.
Point Claude Code or Cursor at AppGram over MCP and the assistant that already has your product open drafts the survey — branching, ready to edit, with every answer attached to the customer who gave it.
Book a demo →Built by prompt
Describe it over MCP or in-app. Branches and scales come with it.
Edited by hand
A generated survey is a normal survey — every question stays editable.
Rates, not just counts
Response rate, completion rate, and average rating, tracked over time.
GENERATE
Describe it. Get a working draft.
Build with AI takes a sentence and returns the questions, the branch conditions and the rating scales already wired. It validates the draft before saving and creates it disabled, so nothing reaches a customer until you have read it.
- Build with AI in the dashboard, or the same tools over MCP from Claude Code or Cursor
- Built-in templates for product/market fit, CES and CSAT — install and fill in the blanks
- Branch conditions inferred from your description, not clicked through a wizard
- Every draft validated against the schema, and created disabled
EDIT
Change anything it got wrong.
A generated survey is an ordinary survey — the same cards on the same Visual Logic Canvas you would have dragged out by hand. Reword a question, swap a rating for a multiple choice, or drag one card's dot onto another and the line carries the condition.
- Six question types — rating, yes/no, multiple choice, checkboxes, short answer, paragraph
- Every question editable: wording, type, scale, required or optional
- Drag to connect — the labelled line between two cards is the branch
- Edit with AI for bulk changes: "make question 2 multiple choice with four options"
ASK
Ask in the product, not in an inbox.
The API returns only the surveys a given customer still owes, so the question goes in your own UI while the answer is fresh — as onboarding ends, as a ticket closes. When a link is the better tool, the same survey runs hosted on your domain.
- Per-customer targeting — the API returns only what that person has not answered
- Hosted run on your portal domain, themed to match your product
- Fingerprinted responses, so a refresh never counts twice
- Metadata at submit time — the plan, the screen, the ticket it followed
READ
See exactly where people drop off.
Question Performance puts a response rate on every question, so the one that loses half your respondents is obvious rather than buried in a total. Under it sits every individual response, carrying the customer and the metadata you passed.
- A response rate per question, not just a total at the top
- Answer distribution per option, so a multiple choice reads at a glance
- Individual responses, each carrying its customer and metadata
- Export CSV in one click, or read the same data through the API
WHO USES IT
What teams actually do with Surveys.
For support leads
Run CSAT with the ticket attached
A rating question fires the moment a ticket closes. A 2 arrives with the ticket, the agent and the customer, so the follow-up is a conversation instead of a number in a report.
For growth
Catch promoters while it is warm
A 9 or a 10 branches straight into a request to quote them, while the goodwill is still fresh — instead of a cold email asking for a testimonial four months later.
CONNECTED
Surveys plugs into everything else you run.
Every module writes to the same customer graph, so data captured here shows up wherever it is useful — no connectors to authorize, no sync to babysit.
See all integrations →AppGram modules
FAQ
What teams ask about Surveys.
How does building a survey over MCP work?
AppGram's MCP server exposes the same backend the dashboard talks to, authenticated with an API key from your organization. Connect it to Claude Code, Claude Desktop, Cursor, or any MCP client and you can describe a survey in plain language; the assistant drafts it, validates it against the schema, creates it disabled, and previews which path a given answer would take. Key scope decides what it can reach — a project-scoped key only ever sees that project.
Do we have to use AI to build a survey?
No. The generator is a faster first draft, not a requirement — the editor is the same editor either way, and starting from a blank survey is a perfectly normal thing to do. Nothing about a generated survey is different from one you built question by question.
What question types can we ask?
Six: rating with a range you set, yes/no, multiple choice, checkboxes, short answer, and paragraph. Each one can be required or optional, and any of them can branch to a different next question.
How do we run NPS or CSAT?
Both are a rating question with a range — 0 to 10 for NPS, 1 to 5 for CSAT. Because ratings branch, the follow-up differs by score: detractors get asked what went wrong, promoters get asked for a quote, and the two never see each other's question.
How does a survey reach a customer?
Two ways, same survey. Ask the API which surveys a given customer still owes and render it inside your own product, or send them the hosted survey link on your portal domain. Responses land in the same place either way.
Can someone answer twice?
Every response carries a fingerprint, so a refresh or a second tab does not become a second answer. Responses from signed-in customers are tied to the customer id you pass, so the answer sits on their record rather than in an anonymous pile.
Can we get the raw responses out?
Yes. Export a survey's responses in one call, or read them through the API one at a time with the metadata you attached at submit time. Nothing is locked in a dashboard.
GET STARTED
Stop paying for six tools. Run product ops from one.
Everything you were bolting together — feedback, roadmap, changelog, help, support, status, forms — in a single workspace with one API and one bill.
Cancel anytime. Yours to test on real customers.