Turn user insight into shipped features

Palette is an agentic research and product-development platform that gathers user insight, validates what to build, and turns findings into shipped features. Get on the waitlist →

“I want to try checkout before I create an account.”

User study

Jul 17 · 10:42

Checkout abandoned at the account form.

Usability study

Jul 17 · 10:39

“Why can’t I pay as a guest?”

Microstudy

Jul 17 · 10:31

User returned to the cart twice, then left.

Screen observer

Jul 17 · 10:27

“Sign-up should happen after I buy.”

User voice

Jul 17 · 10:18

Account creation blocks the payment step.

Slack

Jul 17 · 10:03

8 of 10 participants stopped at the same form.

User study

Jul 17 · 09:54

“I didn’t want another account.”

Usability study

Jul 17 · 09:47

Guest checkout selected after retrying payment.

Microstudy

Jul 17 · 09:39

Cart completion fell at account creation.

Screen observer

Jul 17 · 09:22

“The form made a simple purchase feel permanent.”

User voice

Jul 17 · 09:11

Payment resumed when the account gate was removed.

Linear

Jul 17 · 08:58

“Let me buy first and save my details later.”

User study

Jul 17 · 08:41

Account-required checkout mentioned in support.

Usability study

Jul 17 · 08:34

“I nearly quit when checkout asked me to register.”

Microstudy

Jul 17 · 08:21

Account gate opened after delivery details.

Screen observer

Jul 17 · 08:12

“I only needed to make one purchase.”

User voice

Jul 17 · 08:04

Third checkout retry ended at sign-up.

Google Drive

Jul 17 · 07:51

“Guest checkout was hidden behind login.”

User study

Jul 17 · 07:43

Support thread cites the account requirement.

Usability study

Jul 17 · 07:31

Purchase completed after switching devices.

Microstudy

Jul 17 · 07:18

“I expected payment, not onboarding.”

Screen observer

Jul 17 · 07:07

Account creation added 42 seconds to checkout.

User voice

Jul 17 · 06:55

Session ended before password creation.

Jira

Jul 17 · 06:46

“Save my order before asking me to join.”

User study

Jul 17 · 06:37

Returning buyer chose guest checkout.

Usability study

Jul 17 · 06:29

Checkout exit followed email validation.

Microstudy

Jul 17 · 06:14

“I wasn’t ready to share more details.”

Screen observer

Jul 17 · 06:02

Payment remained open after sign-up exit.

User voice

Jul 17 · 05:48

Guest path completed on the first attempt.

HubSpot

Jul 17 · 05:37

Flow

Autonomous by design. Fully adjustable to your workflow.

Insight

Opportunities

Shape

Compare

Build

Insight

Opportunities

Shape

Build

Compare

Palette collects user insight on its own

Palette collects user insight autonomously and continuously, so your workspace always has the most current picture of what users want and think about your product. Every insight is backed by a real source and weighted by relevance before being added to the workspace's evidence graph.

How Palette collects insight

User studies

Palette creates, recruits, and conducts AI-moderated interviews with real people for in-depth insight on your product, market, and competitors. The interviews are dynamic: Palette asks follow-up questions to surface the insight that only comes forward in conversation.

Usability studies

Palette gives users a goal on your real product and studies how they reach it. The results are parsed, analyzed, and converted into insight.

Microstudies

Palette identifies what it needs to learn about your product, writes questions, and places them on your site for users to answer: autonomously and always with your approval.

Integrations

Palette plugs into the tools your team already uses and reads the user signal that lands in them.

Screen observer

Always on by default. Palette watches every session of your product in use, finds the friction, and collects it as insight.

User voice

Palette launches armies of subagents to read the latest community threads, review sites, social posts, and more to understand what your users say.

Users talked to today

1,284

Signals considered across 146 sources

Opportunities identified

Opportunities being developed

Insights are understood by Palette and turned into feature proposals

Palette groups related user insights that tell the same story and turns them into feature proposals. Each proposal is ranked by the strength and relevance of its supporting evidence, with every insight and source fully transparent to your team. The result is a consistent, evidence-backed ranking of the features your users want most.

Opportunities

“Search should understand the words I use.”12 35 58 81 04 27 50 73 96 19 42 65 88 11 34 57 80 03 26 49 72 95 18 41 64 87 10 33 56 79 02 25

84 58 32 06 80 54 28 02 76 50 24 98 72 46 20 94 68 42 16 90 64 38 12 86 60 34 08 82 56 30 04 78 52 26 00 74 48 22 96 70

“The dashboard takes too long to scan.”49 72 95 18 41 64 87 10 33 56 79 02 25 48 71 94 17 40 63 86 09 32 55 78 01 24 47 70 93 16 39 62

54 28 02 76 50 24 98 72 46 20 94 68 42 16 90 64 38 12 86 60 34 08 82 56 30 04 78 52 26 00 74 48 22 96 70 44 18 92 66 40

“I missed the pricing limit until it blocked me.”86 09 32 55 78 01 24 47 70 93 16 39 62 85 08 31 54 77 00 23 46 69 92 15 38 61 84 07 30 53 76 99

24 98 72 46 20 94 68 42 16 90 64 38 12 86 60 34 08 82 56 30 04 78 52 26 00 74 48 22 96 70 44 18 92 66 40 14 88 62 36 10

“Let me invite the team after setup.”23 46 69 92 15 38 61 84 07 30 53 76 99 22 45 68 91 14 37 60 83 06 29 52 75 98 21 44 67 90 13 36

94 68 42 16 90 64 38 12 86 60 34 08 82 56 30 04 78 52 26 00 74 48 22 96 70 44 18 92 66 40 14 88 62 36 10 84 58 32 06 80

“Notifications keep pulling me back.”60 83 06 29 52 75 98 21 44 67 90 13 36 59 82 05 28 51 74 97 20 43 66 89 12 35 58 81 04 27 50 73

64 38 12 86 60 34 08 82 56 30 04 78 52 26 00 74 48 22 96 70 44 18 92 66 40 14 88 62 36 10 84 58 32 06 80 54 28 02 76 50

“The mobile menu hides my recent work.”97 20 43 66 89 12 35 58 81 04 27 50 73 96 19 42 65 88 11 34 57 80 03 26 49 72 95 18 41 64 87 10

34 08 82 56 30 04 78 52 26 00 74 48 22 96 70 44 18 92 66 40 14 88 62 36 10 84 58 32 06 80 54 28 02 76 50 24 98 72 46 20

“Exporting a report takes three tries.”34 57 80 03 26 49 72 95 18 41 64 87 10 33 56 79 02 25 48 71 94 17 40 63 86 09 32 55 78 01 24 47

04 78 52 26 00 74 48 22 96 70 44 18 92 66 40 14 88 62 36 10 84 58 32 06 80 54 28 02 76 50 24 98 72 46 20 94 68 42 16 90

“I expected the filter to remember me.”71 94 17 40 63 86 09 32 55 78 01 24 47 70 93 16 39 62 85 08 31 54 77 00 23 46 69 92 15 38 61 84

74 48 22 96 70 44 18 92 66 40 14 88 62 36 10 84 58 32 06 80 54 28 02 76 50 24 98 72 46 20 94 68 42 16 90 64 38 12 86 60

“The error says what failed, not how to fix it.”08 31 54 77 00 23 46 69 92 15 38 61 84 07 30 53 76 99 22 45 68 91 14 37 60 83 06 29 52 75 98 21

44 18 92 66 40 14 88 62 36 10 84 58 32 06 80 54 28 02 76 50 24 98 72 46 20 94 68 42 16 90 64 38 12 86 60 34 08 82 56 30

“Onboarding asks for too much at once.”45 68 91 14 37 60 83 06 29 52 75 98 21 44 67 90 13 36 59 82 05 28 51 74 97 20 43 66 89 12 35 58

14 88 62 36 10 84 58 32 06 80 54 28 02 76 50 24 98 72 46 20 94 68 42 16 90 64 38 12 86 60 34 08 82 56 30 04 78 52 26 00

“I want to try checkout before I create an account.”82 05 28 51 74 97 20 43 66 89 12 35 58 81 04 27 50 73 96 19 42 65 88 11 34 57 80 03 26 49 72 95

84 58 32 06 80 54 28 02 76 50 24 98 72 46 20 94 68 42 16 90 64 38 12 86 60 34 08 82 56 30 04 78 52 26 00 74 48 22 96 70

“The search results feel randomly ordered.”19 42 65 88 11 34 57 80 03 26 49 72 95 18 41 64 87 10 33 56 79 02 25 48 71 94 17 40 63 86 09 32

54 28 02 76 50 24 98 72 46 20 94 68 42 16 90 64 38 12 86 60 34 08 82 56 30 04 78 52 26 00 74 48 22 96 70 44 18 92 66 40

“Pricing should explain what counts as a seat.”56 79 02 25 48 71 94 17 40 63 86 09 32 55 78 01 24 47 70 93 16 39 62 85 08 31 54 77 00 23 46 69

24 98 72 46 20 94 68 42 16 90 64 38 12 86 60 34 08 82 56 30 04 78 52 26 00 74 48 22 96 70 44 18 92 66 40 14 88 62 36 10

“I can’t find drafts on mobile.”93 16 39 62 85 08 31 54 77 00 23 46 69 92 15 38 61 84 07 30 53 76 99 22 45 68 91 14 37 60 83 06

94 68 42 16 90 64 38 12 86 60 34 08 82 56 30 04 78 52 26 00 74 48 22 96 70 44 18 92 66 40 14 88 62 36 10 84 58 32 06 80

“I need a calmer notification digest.”30 53 76 99 22 45 68 91 14 37 60 83 06 29 52 75 98 21 44 67 90 13 36 59 82 05 28 51 74 97 20 43

64 38 12 86 60 34 08 82 56 30 04 78 52 26 00 74 48 22 96 70 44 18 92 66 40 14 88 62 36 10 84 58 32 06 80 54 28 02 76 50

“The dashboard should open where I left it.”67 90 13 36 59 82 05 28 51 74 97 20 43 66 89 12 35 58 81 04 27 50 73 96 19 42 65 88 11 34 57 80

34 08 82 56 30 04 78 52 26 00 74 48 22 96 70 44 18 92 66 40 14 88 62 36 10 84 58 32 06 80 54 28 02 76 50 24 98 72 46 20

“Bulk export stalls on larger projects.”04 27 50 73 96 19 42 65 88 11 34 57 80 03 26 49 72 95 18 41 64 87 10 33 56 79 02 25 48 71 94 17

04 78 52 26 00 74 48 22 96 70 44 18 92 66 40 14 88 62 36 10 84 58 32 06 80 54 28 02 76 50 24 98 72 46 20 94 68 42 16 90

“Let me skip templates and start blank.”41 64 87 10 33 56 79 02 25 48 71 94 17 40 63 86 09 32 55 78 01 24 47 70 93 16 39 62 85 08 31 54

74 48 22 96 70 44 18 92 66 40 14 88 62 36 10 84 58 32 06 80 54 28 02 76 50 24 98 72 46 20 94 68 42 16 90 64 38 12 86 60

“The empty state doesn’t tell me what comes next.”78 01 24 47 70 93 16 39 62 85 08 31 54 77 00 23 46 69 92 15 38 61 84 07 30 53 76 99 22 45 68 91

44 18 92 66 40 14 88 62 36 10 84 58 32 06 80 54 28 02 76 50 24 98 72 46 20 94 68 42 16 90 64 38 12 86 60 34 08 82 56 30

“Guest checkout was hidden behind login.”15 38 61 84 07 30 53 76 99 22 45 68 91 14 37 60 83 06 29 52 75 98 21 44 67 90 13 36 59 82 05 28

14 88 62 36 10 84 58 32 06 80 54 28 02 76 50 24 98 72 46 20 94 68 42 16 90 64 38 12 86 60 34 08 82 56 30 04 78 52 26 00

++-.-11*-:*0:-0:*-*-*+000*-:+--..-.-++0:1++::11-:.::00-:0.:+.--+.-0-.+-+----++-::-:--+1.:-:*++-+.+.-:-*-.+.+0---0:*++:.-.:-1*-:+0-1-0*110+0+0+-+011:::::+-:1-*-:+*++:1:-*:1+1--1:-*+--+11.--.0.:.+:*+.:-:*+:-:+-*-*-:::.:.-*+10:1--+-00*-:-+--*1-.::--++0:0+-*-*::+1-++::.*1011-0::.++0:+*+.:+0-.*-+-1:0:0**+++--1-:0-+-:++-:+-*0*::++::+.:.:1.:*1-.1:*:-1:+0--..+1-:-.:-+:1*1*+---010+::*+1.:--*-.:0.0-1.+:-:-:..-.::++..0*-*-+:-+*0::U.s:e0r:s- +g0i*v:e+ :u+p- -a-t. :c:h*e:c.k:o-u:t:..*+:*:0:+:-+-:++0*1101:++.1:+1:+-:--:++:-.::::::+--*1.-*+-1++10.::::------:-:1:*1-::::.:.++:-01::-1--0.*-1-1..++0+++:.:+*01.0-+.-0+-.+0-*:1001.-..1:+:-.++*+1:-+:-1++1+*+0::.1-:..*0-+:1.+++*++-:1::+100:-*.--+-+:+1:-+:01++:+*1*-::-:*---+++-:+1*-+0+00-+1+-:++1-+1-:::+.001-+*--:+.1+1::++00**++-:-0++:**.-*1:0-::1.++::-.-:.:++++*--*:1---::-+0.:-:0:+--:-::.10:100-::+:-0-0:0+:-1-:1.-+00:0+:+--.-+.*+:+0-:::1-.1-:.0-+-*.-1111.:-

Simpler checkout78

0.91384271

−0.781245

0.084031

0.9600047

0.75192834

−0.088142

0.821406

0.94278153

−0.7300441

0.86612

0.84000396

0.791845

0.9230417

−0.077284

0.85019384

0.800004

0.9514286

−0.742013

0.89057142

0.831906

0.9000081

−0.076842

0.87200473

0.810319

0.93000058

−0.721458

0.84719302

0.840071

0.8399964

0.84000001

0.290138014

0.166247579

0.1000000

0.12600004

0.0701842

−0.05021984

0.702385593

0.90000013

0.78000000

−0.31877142

0.440000003

0.2183947561

−0.00781492

0.605731884

0.472908116

−0.193604827

0.882461937

−0.06420871

0.915730004

0.76348219

0.845000021

−0.1129634

0.798421506

0.934760001

0.68824139

−0.04137582

0.907162845

0.824000006

0.756391204

0.88947163

−0.093604218

0.861250009

Smarter search63

Faster onboarding55

Clearer pricing49

Fewer notifications41

Better mobile nav34

Shorter forms28

Cleaner dashboard23

Quicker exports19

Calmer error states15

More useful filters13

Clearer empty states12

Easier team invites11

Safer permissions10

Better file previews9

Faster bulk edits8

Smoother handoffs7

Fewer duplicate alerts6

Clearer settings5

Converting a proposal to a spec

Palette turns a proposed feature into a build-ready specification by considering all the context in your workspace, including previous feature specifications, research, and studies. Based on this context, Palette develops a spec draft, and then validates its hypotheses with users, builds prototypes, runs usability tests, and uses the resulting evidence to keep the feature aligned with the user demand that prompted it.

Shape

Every step is recorded in Palette's decision graph, where each node shows Palette’s reasoning, the evidence collected, and the decisions made along the way. The process is designed for collaboration: Palette can gather and synthesize far more context in far less time, while people contribute judgment and product understanding Palette may not have. Palette gathers the evidence, presents the context, and proposes decisions; your team can review, guide, or make those decisions as the specification develops. It is a new and better way to develop features, one that considers more context and makes every decision easy to trace back to real evidence. If you want Palette to work fully autonomously in the feature spec development process, you can specify so in the workspace settings.

Palette

8 of 10 participants stopped at the account form and didn't proceed with checkout. I suggest we proceed with studying how adjacent companies handle checkout flows to see how they differ from ours in order to understand the low checkout success rate.

All 10 were asked to complete the purchase in their cart.

Participant 1

Participant 2

Participant 3

Usability test completed in 6 minutes 12 seconds. Abandoned the flow when checkout demanded an account.

Participant 4

Participant 5

Participant 6

With our current implementation, the customer must have an account to checkout. I suggest we add a 'checkout as a guest' button allowing non-users to checkout.

Noted. I'll record that in the feature spec.

Steer Palette

Turning a spec into a PR

Palette turns the feature spec developed in Shape into an implementation task that it executes. Any feature spec can be sent to your favorite workflow tool instead of being built in Palette.

You can join the build by prompting Palette as you do with your existing coding tools, or let Palette run by itself. Palette can automatically open and merge PRs for a fully autonomous workflow.

import { type FormEvent, useState } from "react";

import { useRouter } from "next/navigation";

import { checkout, type Cart, type Order } from "@/lib/checkout";

import { GuestToggle } from "./guest-toggle";

import { ReceiptEmailField } from "./receipt-email-field";

import { SubmitButton } from "./submit-button";

type Props = {

cart: Cart;

onComplete?: (order: Order) => void;

};

type SubmitState = "idle" | "building" | "complete";

function receiptEmail(cart: Cart) {

return cart.customer?.email ?? "";

}

export function GuestCheckout({ cart, onComplete }: Props) {

const router = useRouter();

const [asGuest, setAsGuest] = useState(true);

const [state, setState] = useState<SubmitState>("idle");

const [error, setError] = useState("");

async function createSession() {

if (asGuest) {

return checkout.createGuestSession(cart);

}

return checkout.createAccountSession(cart);

}

async function submit(event: FormEvent<HTMLFormElement>) {

event.preventDefault();

setState("building");

setError("");

try {

const session = await createSession();

const order = await checkout.confirm(session.id);

onComplete?.(order);

router.push(`/checkout/confirm/${order.id}`);

} catch (cause) {

setError(checkout.messageFor(cause));

setState("idle");

}

return (

<GuestToggle

value={asGuest}

onChange={setAsGuest}

disabled={state === "building"}

<ReceiptEmailField

defaultValue={receiptEmail(cart)}

disabled={state === "building"}

{error ? <CheckoutError>{error}</CheckoutError> : null}

{state === "building" ? "Creating checkout..." : "Checkout"}

</SubmitButton>

</form>

);

}

Build

Compare

Palette compares before and after

After a feature ships, Palette measures its impact against the user insight that prompted it. Palette learns from the outcome, improving how it develops future feature specifications and recommends what to build next.

3 tasks

Add a guest checkout option to the payment step.
Skip account creation for guest checkout.
Style the checkout button to match the primary action.

Make the button black, our inverted color.

Updated. The checkout button now uses the inverted fill.

Request a change…

OriginalParsed value

window.duration.days14.000000000000

payment.sessions.matched4812.000000000000

payment_step.completion0.610284719302

account_step.abandonment0.341028461293

purchase.duration.normalized0.581842210930

guest_route.exposure0.000000000000

checkout_api.success0.782410036925

guest_session.creation0.000000000000

receipt_email.completion0.603319427506

source_diversity.weight0.681240057193

evidence.confidence0.840001320771

evidence.recency0.702184259842

missing_evidence.weight0.290138014441

contradiction.penalty-0.126000040003

matched_window.confidence0.917042568021

outcome.rank_weight0.635079003118

NewParsed value

window.duration.days14.000000000000

payment.sessions.matched4965.000000000000

payment_step.completion0.821739104148

account_step.abandonment0.089733182405

purchase.duration.normalized0.390720184410

guest_route.exposure0.839996400218

checkout_api.success0.982661420042

guest_session.creation0.961840227519

receipt_email.completion0.874402190067

source_diversity.weight0.702385593104

evidence.confidence0.900000130058

evidence.recency0.780000000019

missing_evidence.weight0.218394756103

contradiction.penalty-0.050219840067

matched_window.confidence0.993158610437

outcome.rank_weight0.839996400218

Palette compares before and after

Old version

OriginalParsed value

window.duration.days14.000000000000

payment.sessions.matched4812.000000000000

payment_step.completion0.610284719302

account_step.abandonment0.341028461293

purchase.duration.normalized0.581842210930

guest_route.exposure0.000000000000

checkout_api.success0.782410036925

guest_session.creation0.000000000000

receipt_email.completion0.603319427506

source_diversity.weight0.681240057193

evidence.confidence0.840001320771

evidence.recency0.702184259842

missing_evidence.weight0.290138014441

contradiction.penalty-0.126000040003

matched_window.confidence0.917042568021

outcome.rank_weight0.635079003118

New version

NewParsed value

window.duration.days14.000000000000

payment.sessions.matched4965.000000000000

payment_step.completion0.821739104148

account_step.abandonment0.089733182405

purchase.duration.normalized0.390720184410

guest_route.exposure0.839996400218

checkout_api.success0.982661420042

guest_session.creation0.961840227519

receipt_email.completion0.874402190067

source_diversity.weight0.702385593104

evidence.confidence0.900000130058

evidence.recency0.780000000019

missing_evidence.weight0.218394756103

contradiction.penalty-0.050219840067

matched_window.confidence0.993158610437

outcome.rank_weight0.839996400218

Self-improving. Better features.

Palette works wherever you do

Access your user insight in Palette from anywhere. Ask what to build, bring insight into your code, or launch a study from Slack.

Claude Code / Codex / Cursor

~/acme/checkout

feature/cart-persistence

Calledpalette_mcp.search_insights

Found recurring cross-device cart loss.

Editedapp/checkout/cart-session.ts

const customer = await getCustomer();

- const cart = createCart();

+ const cart = await restoreCart(customer.id);

+ await persistCart(cart);

Calledpalette_mcp.get_evidence

Linked supporting evidence to the change.

Ranpnpm test checkout

Tests passed

Connect to the Palette MCP server →

Slack

Acme / # product-research

Anne / 9:39 AM

Do we know why younger customers abandon checkout after switching devices?

Alex / 9:41 AM

@Palette launch a study with customers aged 18–24 who switched devices during checkout.

Palette / 9:41 AM

Study created. Currently recruiting 30 participants. I'll let you know once the responses are in.

Set up Palette in Slack →

The setup, in full.

GitHubNot connected

</main>

</body>

Frequently asked questions

What is Palette?: Palette is an agentic research and product-development platform that gathers user insight, validates what to build, and turns findings into shipped features. From insight to build, in one place.
How does Palette gather user insights?: Palette does the research for you. It autonomously runs microstudies, user studies, and usability studies, and analyzes community conversations and product usage sessions. The research is done continuously, keeping your team up to date with what users say and how they use your product. Palette can also connect to the tools where your existing user signals live, including Slack, Linear, Jira, Google Drive, Figma, GitHub, and HubSpot.
How much does Palette automate, and how much do I do?: Palette is designed to work autonomously, from gathering user insights to building features. You choose where it pauses for approval or feedback. For example, Palette can gather insights on its own and bring you in when feature development begins. Every part of Palette's workflow can be customized in settings to fit how your team works.
Can I use Palette outside the web app?: Yes. Palette works in Slack and through its MCP server for Claude Code, Codex, and Cursor. This lets you and your team access your workspace’s user insights, launch studies, and ask Palette for guidance throughout the product development process, all from the tools where you already work.
How do pricing and credits work?: Anyone can sign up and use Palette for free. Free includes 500 credits for your first workspace. When you’re ready to bring in your team, you can upgrade your workspace to Pro for $20 per member per month, with unlimited collaborators. Palette uses workspace credits for actions such as launching user studies and usability studies or building features. You can purchase credits at any time, and they never expire. Palette shows the cost before every credit-spending action and asks for your approval. You can turn these approval prompts off or specify a price range in which they are automatically approved in workspace settings for a more autonomous workflow.
How do I get access?: Join the waitlist with your email. Palette is inviting people in waves, and we’ll email you when it’s your turn to access Palette.

Turn user insight into shipped features

Autonomous by design. Fully adjustable to your workflow.

How Palette collects insight

Your company brain for user insight.

Insights are understood by Palette and turned into feature proposals

Converting a proposal to a spec

Turning a spec into a PR

Palette compares before and after

Palette compares before and after

Self-improving. Better features.

Palette works wherever you do

Claude Code / Codex / Cursor

Slack

The setup, in full.

Frequently asked questions