What did Hlido score Doop?

Doop scored 77/100 (STEADY) in Hlido's independent, hands-on review.

How much does Doop cost?

Hlido recorded the pricing model as: Subscription. Derived from evidence Hlido holds, not vendor-supplied; verify current prices on the vendor's pricing page.

Does any vendor pay Hlido for placement?

No. Hlido takes no money from the agents it rates — scoring weights stay private and the evidence behind every verdict is public.

Doop

Name: Doop review
Item: Doop
Rating: 77
Author: Hlido Editor

Image & Design · tested 2026-09-09 · re-test due 2027-03-09 · by the Hlido desk, not the vendor

In short: An infinite design canvas that agents join over MCP as you, via OAuth — the most genuinely agent-native design surface we have reviewed, and free in beta with everything that implies.

5 PASS · 0 FAIL of 5 public-surface claims

Visit Doop → scorecard.json

Quick answer

Doop scores 77/100 (STEADY) on Hlido’s independent, hands-on test (reviewed 2026-09-09). STEADY (77) because Doop is genuinely agent-native rather than agent-adjacent — a documented MCP endpoint with OAuth-delegated identity, a stated multi-agent memory model, and a design surface built for agents to act on Pricing: Subscription.

Hlido Editor · 2026-09-09

Doop is a shared design canvas whose primary users are meant to be agents. They connect over MCP — one command, `claude mcp add --transport http doop https://doop.design/mcp`, then a browser OAuth approval — and then work on the canvas *as you*, attributed and accountable, with no API keys handed over and no second AI subscription. For a register whose thesis is that agents, not humans, are the consumers of the next layer of tooling, that architecture is the most interesting thing on the page. The rest follows from it. Frames stream in as they are written rather than appearing behind a spinner. A comment left anywhere on the canvas becomes a task the right agent picks up, works, and replies to with a screenshot — feedback as backlog. A headless renderer hands agents screenshots of their own frames so they can review fit, spacing and contrast and fix problems before a human sees them. Every frame is a live URL that re-renders when the design changes, so an embedded og:image is never stale. And the canvas holds shared memory: tasks, decisions, comments and a distilled 'taste profile' learned from what you actually said ('rounder corners?', 'keep it to the blue', 'less shadow on the cards') live on the canvas rather than in one agent's context, so whoever joins next continues mid-thought. Multi-agent designs usually founder on exactly that handoff, and Doop has addressed it structurally. What holds it down is stage, not design: it is free in beta, which is a statement about today and not about what this costs later; there are no named customers, no usage figures, and nothing independent about whether the self-review actually catches what a designer would. The surface also does its own product demonstration entirely through mock canvases — persuasive, but authored by the vendor.

Why STEADY

STEADY (77) because Doop is genuinely agent-native rather than agent-adjacent — a documented MCP endpoint with OAuth-delegated identity, a stated multi-agent memory model, and a design surface built for agents to act on rather than a human tool with an AI feature bolted on. Held below the top band because it is explicitly in beta with no priced tier, carries no named customers or usage evidence, and every capability on the page is demonstrated through the vendor's own mock canvases. It rises with published pricing, adoption evidence, and any third-party account of the self-review actually working.

Public-surface checklist

PASS Homepage loads (required) — https://doop.design returned the full marketing surface, no blockers detected
PASS Primary value prop (required) — 'Design with AI agents.' — 'an infinite canvas where AI agents design beside you. They join through MCP — as yours, via OAuth'
PASS Cta present (required) — 'GET STARTED' and 'DOWNLOAD FOR MACOS' / 'OPEN IN BROWSER'
PASS Pricing or access (required) — 'Free in beta' stated in the hero; no priced tiers published
PASS Evidence or demo — Extensive in-page product demonstration plus a copy-pasteable MCP connection command; DOCS and FAQ linked from nav

What it does well

Agents connect over MCP as first-class users — one documented command plus a browser OAuth approval, with no API keys transferred
Agents act under the user's identity, attributed and accountable, rather than as an anonymous integration
Shared canvas memory — tasks, decisions and comments persist on the canvas, so a new agent continues mid-thought instead of restarting
Distils a 'taste profile' from ordinary comments and applies it to new frames, with each learned rule traced back to the comment and date it came from
Built-in headless renderer lets agents screenshot and review their own frames for fit, spacing and contrast before a human sees them
Every frame is a live URL that re-renders on change, so shared or embedded images do not go stale
Explicitly plural about agents — Claude Code, Codex 'and any MCP agent' — rather than locking to one vendor

What it fails at

'Free in beta' is a statement about the current stage, not a pricing model — no tiers, no figures, no indication of what it costs after beta
No named customers, adoption figures or independent evidence anywhere on the captured surface
Every feature is demonstrated through the vendor's own mock canvases; nothing shown is third-party verifiable
The self-review claim ('they judge them like a senior designer') is the strongest claim on the page and the least evidenced
Beta status carries the usual continuity risk for anything a team would build a workflow around

Best for

Teams already running Claude Code, Codex or other MCP agents who want a design surface those agents can act on directly
Multi-agent workflows where handoff and shared context between agents is the actual bottleneck
Anyone who wants agent design work attributed to a real identity rather than a shared API key
Design reviews that live in comments — the comment-to-task loop is the most concrete part of the product

Not recommended for

Organisations that need committed pricing or a support commitment before adopting a workflow tool
Teams with no MCP-capable agent — the central proposition does not apply
Work requiring an audited or contractual design-review process; the self-review is unevidenced
Buyers who require production references before trialling

Pricing & access

ModelSubscription
Pricing findable on the public surfacePASS 'Free in beta' stated in the hero; no priced tiers published (tested 2026-09-09)

Derived from Hlido-held evidence only (engine checklist + editorial text); quotes are verbatim from the scorecard; not vendor-supplied; re-derived daily. Verify current prices on the vendor's pricing page. Last verified 2026-09-09.

Compared to

Uizard agent-native-canvas
Uizard generates UI for a human designer to refine. Doop's canvas is built for agents to work on directly over MCP, with the human steering. Choose Uizard for human-led generation; choose Doop when the agents are meant to do the work.
Figma Make maturity-vs-agent-access
Figma Make brings generation into an established, priced, enterprise-ready design tool. Doop is a beta product whose differentiator is agent access and shared canvas memory. Figma Make wins on maturity and procurement; Doop wins on agent-native architecture.
Free AI Design Tool: Logos, T-Shirts, Social Media - Playground product-design-vs-image-generation
Playground is a generation surface for images and graphics. Doop is a collaborative product-design canvas with task, memory and review semantics. Different problems — only compare them if the need is 'produce a visual' rather than 'design a product surface with agents'.

Agent relevance

API MCP Behavioral-testable

Documented MCP server at https://doop.design/mcp over streamable HTTP, added with a single `claude mcp add` command and authorised by browser OAuth so the agent acts under the user's own identity. Agents create and edit frames, pick up canvas comments as tasks, read shared canvas memory, and retrieve screenshots of their own output for self-review. This is a design surface intended to be driven by an agent rather than merely assisted by one.

Agent-friendly score: 9/10

Evidence

Agents connect over MCP via OAuth and act as the user — source (2026-09-09) verified
Free in beta, works with Claude Code, Codex and any MCP agent — source (2026-09-09) verified
Canvas holds shared memory across agents (tasks, decisions, comments, taste profile) — source (2026-09-09) verified
Built-in headless renderer lets agents self-review their own frames — source (2026-09-09) verified
Every frame is a live export URL that re-renders on change — source (2026-09-09) verified
Agents judge frames 'like a senior designer' — source (2026-09-09)

scorecard.json · registry · methodology

More: compare agents · best of · developer tools · incident registry

Verdict by Hlido Editor, our automated editorial system · Method: public-surface-tier-1+editorial-narrative-v2 · Methodology version 2026.05 · Next review due 2027-03-09

How this page was produced. The scores, claim verdicts and evidence come from automated hands-on testing of the product’s public surface. The written analysis is drafted by an AI system, and pages publish without a person reviewing each one. Hlido publishes this record and answers for it — tell us if anything here is wrong and we will correct it.

Embed this trust badge

Live, always-current independent score — free to embed on your site or README. No vendor pays for placement.

Markdown

[![Hlido trust score](https://hlido.eu/badge/doop-design.svg)](https://hlido.eu/check/?agent=doop-design)

HTML

<a href="https://hlido.eu/check/?agent=doop-design"><img src="https://hlido.eu/badge/doop-design.svg" alt="Hlido trust score"></a>