# Palette

> Palette is an agentic product-development system that continuously gathers user insights, validates them with real users through studies, turns findings into build-ready specs, and builds the specs out, before measuring their impact. Access is by waitlist; invites go out in waves.

Palette is an agentic research system: from insight to build, in one place. It can run autonomously end to end; your team chooses the approval gates, reviews proposals, and decides what ships. Setup is two steps: connect GitHub and add a one-line script tag to your site.

How Palette works, in five stages:

- Insight: Palette collects user insight autonomously and continuously. Every insight is backed by a real source and weighted by relevance before entering the workspace's evidence graph.
- Opportunities: Palette groups related insights that tell the same story into feature proposals, ranked by the strength and relevance of their supporting evidence, with every insight and source transparent to your team.
- Shape: Palette turns a proposal into a build-ready specification using all workspace context, then validates its hypotheses with users, builds prototypes, and runs usability tests. Every step is recorded in Palette's decision graph; Palette proposes, and your team can review, guide, and decide.
- Build: Palette executes the spec as an implementation task. You can prompt it as you do with existing coding tools or let it run by itself, and it can automatically open and merge PRs for a fully autonomous workflow. Any spec can instead be sent to your favorite workflow tool.
- Compare: After a feature ships, Palette measures its impact against the insight that prompted it and uses the outcome to improve future specifications and build recommendations.

Insight comes from six sources:

- User studies: AI-moderated interviews with real people, with dynamic follow-up questions, for in-depth insight on your product, market, and competitors.
- Usability studies: users are given a goal on your real product and Palette studies how they reach it, converting the results into insight.
- Microstudies: questions Palette writes and places on your site for users to answer, always with your approval.
- Integrations: Palette reads the user signal that lands in Slack, Linear, Jira, Google Drive, Figma, GitHub, and HubSpot.
- Screen observer: on by default, it watches sessions of your product in use and collects friction as insight.
- User voice: subagents read community threads, review sites, and social posts to understand what your users say.

Beyond the web app, Palette works from Slack through the @Palette bot, and from coding tools through the Palette MCP (Model Context Protocol) server for Claude Code, Codex, and Cursor. By default, every action that spends credits asks for approval first; this can be turned off in workspace settings for a fully autonomous experience.

Pricing has two tiers; both include 500 credits to your first workspace. Free includes everything in Palette. Pro costs $20 per member per month and adds unlimited collaborators, 500 credits per month to your workspace, and Palette in Slack. Credits are debited from the workspace balance, are spendable by any workspace member, cover actions such as launching studies and building features, can be purchased at any time, and never expire.

Palette began its public rollout on July 20, 2026. Palette is a product of Isotropic, Inc., based in San Francisco, CA. The product lives at https://palettelabs.ai/. Support: support@palettelabs.ai. Privacy questions: privacy@palettelabs.ai. The home page is very large as served, so on a small context budget prefer the summary above and the full docs text below to fetching it.

## Product

- [Pricing](https://palettelabs.ai/pricing): Free and Pro tiers, both with 500 credits to your first workspace; Pro at $20 per member per month with unlimited collaborators and 500 monthly workspace credits; how credits work.
- [Waitlist](https://palettelabs.ai/waitlist): Leave your email to be notified when it is your turn to access Palette; invites go out in waves.
- [Changelog](https://palettelabs.ai/changelog): What's new in Palette; the first public entry marks the start of the public rollout on July 20, 2026.

## Guides

- [Blog](https://palettelabs.ai/blog): Practical guides for SaaS teams on continuous user research, evidence, feature validation, and product specifications.
- [How SaaS Teams Can Run Continuous User Research](https://palettelabs.ai/blog/how-saas-teams-can-run-continuous-user-research): A sustainable research rhythm that keeps traceable user evidence close to product decisions.
- [AI-Moderated User Interviews: When to Use Them and What Good Evidence Looks Like](https://palettelabs.ai/blog/ai-moderated-user-interviews-good-evidence): When AI moderation fits product research, when a human should lead, and how to judge evidence from real participants.
- [How to Prioritize User Insight](https://palettelabs.ai/blog/prioritize-customer-feedback-using-evidence): How to compare customer problems without treating request totals as a roadmap.
- [How to Validate a Feature Before Building It](https://palettelabs.ai/blog/how-to-validate-a-feature-before-building-it): How to test the problem, solution, and usability before committing engineering time.
- [How to Turn User Research Into a Build-Ready Product Specification](https://palettelabs.ai/blog/turn-user-research-into-build-ready-spec): How to translate validated evidence into scoped, testable product requirements.

## Docs

- [Docs](https://palettelabs.ai/docs): Setup for the @Palette Slack bot and the Palette MCP server, with client configuration for Cursor, VS Code, and Claude Code. One brief page covers both topics.
- [Docs (full text)](https://palettelabs.ai/llms-full.txt): Complete agent-readable Palette documentation from the same Markdown source as the docs page.

## Optional

- [Home](https://palettelabs.ai/): Landing page ("Turn user insight into shipped features") covering the five-stage flow, the six insight sources, autonomy controls, and the Slack and MCP surfaces. Its content is fully summarized above; the page is very large as served.
- [Terms of service](https://palettelabs.ai/legal/terms)
- [Privacy policy](https://palettelabs.ai/privacy)
- [Waitlist privacy notice](https://palettelabs.ai/waitlist/privacy)