Changelog

A new question type, Budget: respondents share a fixed budget between your priorities, or tick projects until the money runs out. And a budget on a question group: give the map questions inside it a price — a bench £2k each, a path £50 a metre, a flower bed £30 a square metre — and respondents place marks until the group's total is spent. A bar shows what is left; Next waits until the rule you set is met.

Why. Budget and precept consultations were being built from ranking questions, which say what matters but not what people would give up. Responses, exports and public results now show the average share per item, the projects that fit the budget together, and where the placed items went.

Sites is a new item in the sidebar. Build a public site for a project from blocks — text, FAQ, a timeline of the phases, pictures, video, documents to download, a map of the project area — and add cards for your surveys: each card follows its survey from "coming soon" to "open" to "closed, see the results". Show what residents said with charts and maps from the results, and announce an open house with registration and a calendar invitation. One survey can appear on any number of sites.

Why. An engagement is longer than one survey: a pilot, the main round, feedback on the proposal, events. Flyers and QR codes need one address that stays the same from the first pilot to the final plan, and until now that page had to be built by hand on another site. Sites are part of Pro; a published site stays online even if the plan ends.

Map styles is a new page in the sidebar, under Library. It holds the five base maps — Streets, Satellite, Topo and two made for printing, Paper A and Paper B — and the ones you make: duplicate a preset, open the editor, change what stays on the map, the colours of land, parks and water, the look of streets and labels, and see the result at once, online and as the printed form will render it. Pick any style in a survey's settings; export it as a file and import it into another workspace.

Why. Planners and consultancies want the survey map to match their own cartography and to stay the same across every survey, and a paper form needs a map that survives a photocopy. Until now that meant asking us for a custom base map. Creating and editing styles is part of Pro; using any style that exists never is, so a survey keeps its map whatever the plan.

Points that sit close together can now merge into one circle with a count. The circle splits into smaller groups and single points as you zoom in or click it. It is easier to read than plain markers when answers are spread far apart, for example research sites around the world.

On Responses → Map, open a point layer's ⋮ menu and choose Create Clusters. The clusters follow your filter, and removing them brings the points back. On a public results page, set a map block's Visualization to Clusters.

Every creator page now has a sidebar on the left. Your workspace is at the top: click its name to switch to another one. Below it are Surveys, API (your tokens), MCP (connecting Claude, ChatGPT and other assistants), and your workspace's Members, Plan & billing and General, one click away instead of inside a menu.

Inside a survey the sidebar shrinks to icons so the map keeps its width. On the survey list you can collapse it too, and it stays the way you left it. On a phone it opens from the ☰ button.

Someone you invite from Workspace → Members now arrives where you invited them. A colleague who creates an account from the invitation link joins your workspace straight away, with no extra workspace of their own, and sees your dashboard with a note that they have joined. Someone who confirms an invited address later lands there too, instead of on a page asking them to start a survey.

An editor who does not yet see any surveys reads why: surveys they create, and the ones you add them to, appear on their dashboard.

Settings → Survey period takes an opening time, a closing time, or both. Before the start respondents see a page saying when the survey opens; at the end it closes by itself, within five minutes, exactly as if you had pressed Close — and the status chip shows what comes next.

Reopening a survey by hand clears a closing time that has passed, so it stays open until you set a new one. The same fields are available over the REST API and the MCP server.

Each section now has a Layout block. Put a picture where the map would be (a welcome page with a photo of the place), move the questions to the right, make them a third or half of the screen, and set them as a calm column in the middle of the panel.

On a map page you can keep the questions open while respondents place a point or draw: the crosshair and the drawing buttons aim at the map beside them, so nobody loses the question they are answering.

Every setting is optional and applies on computer screens; phones keep the familiar layout, and existing surveys look exactly as before.

Every survey now has a Translations tab, between Survey and Responses. It lists every text of the survey once — page titles, questions, help texts, answer options, the Next button, the thank-you page — in the order respondents see it, with the original on the left and the translation on the right. "Other" and "(Select all that apply)" appear once and one translation fills every place they are used; a row can be translated separately when the same words need different wording in context. Each row links back to its question in the editor.

Rows show what is missing and what is outdated — a translation whose original was edited after it was written — and the publish warning counts both. Export the table as a CSV for a translator, import it back, and the report says what was filled, skipped or not found. Languages are added right there, and a second language of the survey can serve as the source column for a translator who does not read the primary one.

The translation fields in the question, section and group dialogs now show the stored text and open its row on the new tab, so there is one place where translations are written.

The third card on New survey is live: Use your AI assistant. Pick yours — Claude, ChatGPT, Gemini, Le Chat, Perplexity, Cursor, VS Code, Claude Code and more — and follow the shortest way it allows. For Claude it is one button: the connector form opens filled in, you press Add, Connect and Allow. The page notices by itself when you are connected.

Then write what you want to find out, open the prompt in your assistant, and the draft appears on your dashboard. If your assistant cannot connect to outside tools yet, the page says so and offers the Survey wizard instead. Building surveys through an assistant is part of Pro.

Mapsurvey now has an MCP server. Connect Claude, ChatGPT, Claude Code, Cursor or any agent that speaks the Model Context Protocol, and ask it in plain words: "create a survey in German and English about where benches are missing, with a point question and an urgency rating". It hands you the link, a QR code for a poster and campaign links to send it out. Once answers come in, ask it to summarise them, read the comments, map what people drew, or give you an Excel file. It sees only your surveys, works through the same rules as the editor, and asks you before it publishes or deletes anything.

How to connect. Every assistant signs you in — no token to copy; scripts can use a personal token made for the MCP server. New survey → Use your AI assistant has the steps for each one, and the help centre has the set-up guide and every tool. The MCP server is a Pro feature: tools on a survey need its workspace on Pro.

Your personal access tokens now open the whole REST API: create a survey, add sections and questions, publish it, and read or export its responses from a script, a reporting pipeline, a QGIS workflow or your own app. It follows the same rules as the editor — wording on a live survey changes in place, structure goes through unpublished changes, and a token never does more than you could in the browser.

Where to start. The help centre has a step-by-step guide and a reference of every endpoint; the machine-readable contract is /api/v1/openapi.json. The API is a Pro feature: calls on a survey need its workspace on Pro.

Your account now has an API & connections page. Create a personal access token there — with the permissions you choose, for one workspace or all, with or without an expiry — and use it from a script, a GIS tool or an AI agent with Authorization: Bearer …. Revoke it with one click.

What it opens. The first endpoint, /api/v1/me, is live today; the REST API and the MCP server that these tokens unlock are next. The API is a Pro feature: tokens work on any plan, calls on a survey need its workspace on Pro.

Between 13 and 22 October, people who opened a survey from a link inside the Facebook or Instagram app on an iPhone could not send their answers: every "Next" came back with a security error. The same change made a few editor actions fail for creators — moving responses to the trash, publishing public results, dismissing these cards. Both are fixed; nothing you saved was lost, but answers those respondents tried to give in that window never reached you.

Why. Surveys embedded on other sites got cookies that work inside a frame, and those cookies were sent to everyone. In-app browsers on iOS refuse them. Embedded surveys now keep their session another way, and everyone else gets ordinary cookies.

Point questions have a new setting: A tap on the map places a mark. Turn it on, and respondents can tap the map to drop a mark straight away instead of pressing the question's button first. The mark opens the same pop-up and counts towards the same limits. The button still works.

The setting only works while the point question is the only map question on its page. If another one is on the page, the question's "tap to place" badge turns grey and taps do nothing.

Why. Respondents often click the map first and nothing happens. Point questions are the questions people skip most.

The Coverage surface on the Responses map now counts people, not shapes: someone who drew two overlapping areas counts once, and the row's "of N respondents" is the number of people.

The strongest spot on the map is now full colour, so agreement shows up even when nowhere near everyone drew the same place. In the coverage settings, Contrast decides how quickly the colour builds up and Min. people hides spots fewer people chose. A small chart under the sliders shows the result for every count.

Why. On real surveys the surface was so pale it was hard to see at all.

Put several questions under one title and description, and show or hide them all with one visibility rule. Click Add group next to New Question, or next to Add Sub-question to group the questions of a map pop-up, then drag questions into it. A group shows as a framed card, a plain heading, or a block that opens on tap.

Why. When twenty questions depended on one answer, the same rule had to be set twenty times, and respondents saw one long list. A group sets it once — for example "Your visit" appears in the pop-up only after the type of place is picked.

Single and multiple choice questions have a new Display as option: Icon grid. Every option becomes a tile with an icon from the map icon set above its label, in the question's colour or a colour of its own. Options that share a group sit together under one heading. On a single choice question, the grid folds to the chosen row once someone picks, and "Show all options" opens it again.

Why. A list of fifteen kinds of place takes two screens to scroll in a pop-up on a phone. As tiles, the whole list fits, and people find their option by its picture.

Creator pages now show a thin bar at the top while the next page or a save is on its way, and a button you pressed turns busy at once: a spinner, and no second send if you press it again.

Why. On a slow connection the Responses page can take a few seconds to arrive, and nothing on screen said the click had registered. People clicked again, and some forms were sent several times in a row.

Publishing Unpublished changes on a live survey no longer resets what you did not touch: conditional questions and sections keep their Visibility rules, form sections stay forms, custom button labels (and their translations) and per-section hidden layers survive the publish.

Why. The working copy behind "Unpublished changes" was created with a list of fields that predated these settings, so every publish quietly put them back to their defaults. If a survey of yours lost its rules or labels on a publish, the previous version still has them: write to us and we restore them, or restore the version yourself and publish it again.

A sub-question of a map question (or of an "Objects on the map" question) now has the same Visibility block as a regular question. Pick an earlier choice sub-question of the same pop-up and the options that should reveal it, and the follow-up appears only for marks where the respondent picked one of them.

One point question with a "Type of place" choice and a follow-up per type now replaces a separate map question per type with the shared follow-ups copied by hand. The rule is decided for every mark separately, a hidden follow-up is never required and never stored, and the section list and the question modal show the same branch badges you know from conditional questions.

+ New Survey now asks how you want to begin: start from scratch (one click, straight into the editor, the map already on your city), the Wizard (one sentence in, a full draft out), or — soon — connect your own AI agent through MCP. Choose your usual way in Preferences (account menu) and the question never comes back; "Start another way…" in the dashboard's ⋯ menu brings it back when you want it.

Why. The create page opened on Berlin for everyone and put the Wizard brief first, whatever you had in mind. Now the map opens where you are when your browser allows it, or where your workspace's last survey was — and the way you like to start is yours to keep.

The hint you write under a question now appears once, small and grey under the question, for every question type. On map questions it used to appear twice, the second copy larger than the question itself.

Sub-questions that open on the map, when a respondent places a point or taps an object, now show their hints too, and ratings there use the style you picked (scale, labeled list or stars) instead of plain radio buttons.

Why. Hints became formatted text, and the old layout could not hold it. In map pop-ups hints were never shown at all.

A ? in the editor header opens a small help menu: what's new, the FAQ, and "Help & support" — a chat with us. If you leave the page before we answer, the reply reaches you by email, and the conversation continues there.

Why. Support used to be hidden in the account menu, and a question asked while we were away was lost: the chat did not know who you were. Now it does — you are signed in, so there is nothing to fill in.

Share now has an Embed on your site section: copy one line of code and the survey runs inside your page, on a city website, a project page or a client portal. Tilda, Webflow, Wix, Squarespace and WordPress all take it; in Notion you just paste the survey link. Responses land here as usual, and Responses → Traffic Sources shows which site they came from.

Forms grow with their content, so no inner scrollbars. For a map, set the height you want in the code. If you only want the survey on your own site, list its address under Sites allowed to embed; leave the list empty for Notion and other platforms that embed from a link.

Why. Creators kept asking to keep respondents on their own pages instead of sending them to a link, and it works everywhere now, including Safari on iPhone.

In Survey settings you now choose what respondents see when they finish: the thank-you page, or a web address you give, such as your project's Facebook page. The address has to be a full link starting with https://, and the field tells you if it is not.

Why. The old "Redirect URL" box accepted anything. A hashtag or a short message typed there kept respondents on the last page, pressing Finish again and again. Surveys that had such a value now show the thank-you page, and a link typed after a "#" works again.

Fixing wording, help text, translations or map settings, or adding an answer option, on a published survey now saves straight to the live survey. There is no draft to create first and no new version.

Adding, removing or reordering questions and sections still waits. Click any of those controls and choose Start unpublished changes. Respondents keep the current survey while you edit. When you are done, click Publish changes.

Why versions exist. Publishing structural changes creates a new version of the survey, so every answer stays tied to the questions that respondent actually saw. Without versions, deleting a question would delete the answers people already gave to it. Changing a question's type or removing an option would leave one column holding answers to two different questions. Responses shows every version together under All versions, and you can still filter and export each version on its own.

Why this change. Many of you opened a published survey to fix a typo and found every field locked. Now only the edits that could break answers you have already collected are locked.

On the Responses map, a polygon or line question now has a Coverage action in its layer menu. It draws, for every spot on the map, how many respondents included it in their area or passed within a set distance of it on their route, so where people agree stands out instead of hiding under a stack of translucent shapes.

The row states the count ("≤ 7 of 12 respondents"); the settings let you change the cell size and, for routes, the corridor width. The surface follows the table selection like a heatmap does.

Why. "Draw the area you consider the centre" and "draw your route" are asked to find agreement, and the map could show it for points and spray but not for shapes.

A respondent can now spray-paint an area instead of drawing a polygon. Pick Spray area in the type picker and choose a default brush; the respondent paints with a mouse or a finger, switches between three brush sizes, erases, and paints thicker where they are more certain. It fits questions whose honest answer has no boundary — where does the centre end, where do you feel unsafe, where is it too hot in summer.

Responses shows every cloud in the question's colour as a grid of agreement: darker cells are where more respondents painted, each counted once however long they held the brush. The same grid goes to the public results page, with cells below the anonymity threshold left out, and never a single respondent's paint. Exports carry each cloud as a MultiPoint with its dot count and painted area. Frozen public pages with a map block will ask you to re-freeze once.

Why. The method comes from PPGIS research (Map-Me, Huck et al. 2014): a hard-edged polygon forces a precision people do not have, so they skip the question or draw too much.

An empty session is someone who opened your survey and left without answering. Responses, the Overview and exports now leave them out. The headline still says how many there were, and Show brings them back.

Why. On widely shared surveys they outnumbered real responses two to one and hid how many people actually took part.

From now on, a card like this one appears once in the editor after an update: what changed, and why. Older updates stay on the What's new page, under the gift icon in the top bar or in the account menu.

"Got it" or the × closes the card. "Don't show these" switches cards off; the page stays, and you can switch them back on there.