Figr
Figr is an AI product design agent for teams working on software that already exists. It learns your screens, flows, design system and documents, then produces UX reviews, user flows, edge cases, PRDs and Figma-ready designs.
What is Figr?
Figr is an AI product design agent built for teams whose product already exists, in the mature and history-laden sense. Its argument is stated bluntly on the homepage: most AI design tools start with a prompt, which works for landing pages and breaks inside real products. So Figr reverses the order. Before it draws anything, it builds a living understanding of the software, a Visual Context Graph assembled from live capture of the running product through a Chrome extension, screen recordings, Figma files, PRDs and research documents, analytics exports and code. That understanding is stored in Context Pods, reusable and shareable memory of product context sized from 200 MB to 3 GB depending on the plan, layered with the team's design system, user feedback and industry benchmarks.
The working unit is a product question rather than a design prompt: what states are missing before engineering starts, how a feature should behave for admins, managers and viewers, what an empty state needs to explain. Figr asks its own clarifying questions, maps the flow that already exists, finds edge cases, compares against known UX patterns, checks accessibility and explains the trade-offs it made. That reasoning is meant to be inspected before any screen is accepted, which is the step most tools skip. Only then does it produce artefacts: user flows, UX reviews, edge-case and test-case maps, information architecture, persona maps, PRDs, handoff documents and high-fidelity prototypes, all constrained by the team's real components and tokens.
Work leaves the tool through two-way Figma sync, PNG or JPG images, React/TSX code, share links, or the Figr MCP server into VS Code and Cursor. The positioning is deliberately non-confrontational, this is not Figma versus Figr: the agent works upstream of the canvas and hands over to it.
The publisher is Figrfast Systems Private Limited, of Bengaluru, with a US entity in Delaware. Anthropic's Claude API is the declared model provider, the company holds SOC 2 Type II audited annually through Sprinto, and it claims more than 500 product teams, with named testimonials from Visma, Letta, Mailerlite, Product School, GreyOrange, Adaptive and ByteXL. Billing runs on credits rather than on seats.
What it does
- Capture a product as it runs, through a Chrome extension, and turn it into a working map of how it actually behaves
- Ingest Figma files, design tokens, CSS and Storybook to learn the team's design system, then hold the output to it
- Take a product question instead of a design prompt, ask its own clarifying questions and map the existing flow first
- Surface missing states, edge cases and role-based variants before a design review or an engineering handoff
- Generate user flows, UX reviews, PRDs, test cases, information architecture, persona maps and high-fidelity prototypes
- Expose its UX reasoning, accessibility checks and trade-offs for inspection before any screen is accepted
- Export to Figma, PNG/JPG or React/TSX, share a link, or feed an IDE through the Figr MCP server
When to use Figr / When not to
A quick filter to help you decide if Figr is the right fit.
When to use Figr
- Product teams maintaining a mature, complex product where every new screen has to fit years of existing behaviour
- Product managers who need a defensible UX direction without waiting for a designer to free up
- Product designers who want generated output constrained by their own tokens and components rather than a generic style
- Teams mapping edge cases, missing states and role-based variants before a design review or an engineering handoff
- High-output individuals and small collaborating teams, the profiles the Starter and Max plans are built around
When not to use Figr
- Anyone starting a landing page or a brand-new product from a blank page: the publisher itself points to prompt-to-screen tools for that job
- Teams hoping to outsource product judgment, which the publisher explicitly says the tool does not remove
- People who work mainly from a phone or a tablet: the interface is optimised for the desktop and there is no mobile app
- Users under 18, excluded by clause 3.1 of the terms
- Organisations handling health or payment data, or needing a contractual accessibility (WCAG) guarantee on the output, both of which the terms rule out
How to use Figr
A typical end-to-end flow, from setup to results.
- Create a free account on the web app; no payment is required to start
- Install the Chrome extension and let Figr capture the product as it runs in production, roughly five minutes according to one customer
- Hand over the starting material: four or five screenshots, Figma tokens, CSS files or a documented design system
- Let Figr auto-explore the flows and configure the design system from its backend, counted at about 20 minutes and up to an hour on a complex product
- Book the optional 15-minute onboarding call if you would rather be walked through the setup
- Feed the everyday context as you go: screenshots, a shared screen, URLs, PDFs, analytics CSV exports
- Ask a product question rather than writing a design prompt, and describe the UX problem you are trying to solve
- Answer the agent's clarifying questions and let it map the flow that already exists
- Inspect the UX reasoning, the missing states and the stated trade-offs before accepting any screen
- Export to Figma, PNG/JPG or React/TSX, share a public or password-protected link, or pull the work into VS Code or Cursor through MCP
Pros & Cons
Pros
- Context ingestion far broader than prompt-to-screen tools: live product capture, Figma, documents, analytics and code
- The UX reasoning is made inspectable before screens are generated, trade-offs included
- Output is constrained by the team's real design system, with tokens, components and versioning
- Round-trip to Figma, plus React/TSX export and an MCP server that reaches into the IDE
- No-training-on-customer-data commitment written into both the terms and the enterprise terms, alongside SOC 2 Type II with annual audit, AES-256 at rest, TLS/HSTS in transit, MFA and a published list of named subprocessors
- A permanent free plan with no expiry, and unlimited read-only guests that cost nothing on any tier
- An unusually detailed plan comparison table on the pricing page, and named customer testimonials with roles and companies
Cons
- No explicit claim of GDPR compliance, no DPO and no EU representative under Article 27
- EU data residency is reserved for the Enterprise plan; Free, Starter and Max default to the United States, and data is processed and stored in India and/or the United States with consent to the transfer requested from the user
- Credit consumption is hard to forecast: the pricing page and its own FAQ publish two different scales for the same tasks, and the credit roll-over rule differs between the terms and the pricing page
- The documentation FAQ still shows an obsolete price grid, with a Pro plan at USD 99 a month and a free plan of 20 one-off credits, contradicting the official pricing page
- Editorial slips are visible on the live site: an internal note asking to add Stripe billing and proration language before launch left on the pricing page, several typos, and footer links for the demo and the cookie policy pointing at nothing
- Paid seats are capped at three below the Enterprise plan, and no uptime or incident-response SLA is offered on Free and Starter
- No mobile app and no announced interface language other than English, and the terms disclaim any accessibility (WCAG) guarantee on the generated output
Pricing & Plans
Figr is sold on a freemium basis. A permanent free plan is available at no cost, with 10 credits and three canvases a month; it is a standing plan rather than a time-limited trial. The cheapest paid entry point is Starter at USD 32 per month on an annual commitment, or USD 39 per month billed monthly. Max costs USD 124 per month annually, or USD 149 monthly, and Enterprise is quoted on request. Annual commitment is presented as saving two months. Usage is billed in credits rather than per seat; subscriptions are charged in advance, renew automatically, and fees are non-refundable unless stated otherwise, with 30 days' notice before any price increase and pro-rata refunds only where the publisher terminates without cause. Prospective buyers should be aware that two different credit-consumption scales are published for the same tasks, so the real monthly cost of a given workload is difficult to estimate in advance.
- 10 credits a month
- 3 canvases
- 1 Context Pod of 200 MB
- limited design system
- export to Figma
- 500 MB of storage
- 24-hour version history
- email support
- 100 credits a month
- unlimited canvases
- 1 fully active seat
- 3 Context Pods of 1 GB
- unlimited read-only guests
- 3 design systems
- custom MCPs
- 5 GB of storage
- 1
- 000 shared credits a month
- up to 3 seats
- unlimited team Context Pods of 3 GB
- unlimited shared design system
- real-time collaboration and comments
- Figr branding removed
- 50 GB of storage
- unlimited seats
- per-seat credit allocation
- SSO with Google
- Okta and custom SAML
- audit logs and activity export
- RBAC
- unlimited Pod storage
- custom MCPs and integrations
Data, GDPR & hosting
A consolidated view of how Figr handles your data.
GDPR overview
Figr never states in so many words that it is GDPR-compliant. Its privacy policy, effective 2 January 2026 and last updated 12 December 2025, does grant people in the EU and EEA the rights of access, rectification, erasure, restriction, portability and objection, with equivalent rights under the UK GDPR, plus detailed CCPA/CPRA and Shine the Light rights for California residents. It describes a complaints procedure with investigation and a written reply, and promises breach notification in line with applicable law. The declared controller is Figrfast Systems Pvt. Ltd. and its affiliates; processing and storage happen in India and/or the United States, with consent to that transfer requested from the user. Two gaps stand out: no EU representative is appointed under Article 27, and no DPO exists. The only named privacy contact is a Grievance Officer under India's Information Technology Act 2000.
Who owns the data?
You keep every ownership right in the data you bring in and in anything you create with the service; Figr receives only a limited, non-exclusive licence, granted so that it can run the product for you. The terms add that Figr claims no ownership over the outputs generated from your data, and the data-protection page puts it in plain words: your data belongs to you, and the competitive edge it carries stays yours. For Service Data the publisher describes itself as a processor acting on your behalf, not as the owner. One exception is worth knowing: any feedback you send is licensed to Figr perpetually, irrevocably, worldwide and free of charge.
Reuse rights
You can reuse your inputs and everything the agent generates for your own product work without asking permission, with a single contractual limit: the terms forbid using the outputs to compete directly with Figr. On the publisher's side, customer data is declared not used to train the AI models, not sold, not shared with other customers and not traded, with technical and organisational isolation behind that commitment; data pulled in through third-party integrations gets the same protection. Two reservations remain. Anonymised, aggregated data may be used to improve the service, for research and to publish performance benchmarks, without identifying customers. And ordinary product analytics apply: Amplitude, Microsoft Clarity and Mixpanel collect usage, session recordings and heatmaps, while Anthropic (Claude API) is the declared model provider. The privacy policy rules out any sale of personal data and any sharing for cross-context behavioural advertising.
Data retention & training
Hosting summary
Figr runs on Amazon Web Services. The security and data-protection pages point to AWS facilities in the United States, while the privacy policy is broader and states that data is processed and stored in India and/or the United States, with consent to that transfer requested from the user; the two statements should be read together. Residency then follows the plan: United States by default on Free, Starter and Max, and United States, EU or a custom region on Enterprise only, which also offers BYOK customer-managed keys and an on-premise deployment. Data is encrypted with AES-256 at rest and travels over TLS with HSTS; AWS sites are credited with 24/7 physical security and access control, and databases are clustered with load-balanced scaling for redundancy. Six subprocessors are named: Amazon Web Services, Anthropic as the declared model provider, Stripe for payments, and Amplitude, Microsoft Clarity and Mixpanel for analytics. SOC 2 Type II is audited annually through Sprinto, with annual penetration tests and a 72-hour incident-notification commitment. The contracting entity depends on the user's location: a Delaware company for the United States, the Indian company elsewhere.
Things to keep in mind
Risks and trade-offs to weigh before adopting Figr.
- Handing over product judgment: the agent returns confident, well-argued flows and PRDs, and the publisher itself says it does not remove the need for judgment. A team that stops challenging the output will ship reasoned-sounding mistakes
- The generated work comes with no guarantee of accuracy, accessibility or regulatory conformity; the terms explicitly disclaim WCAG conformance, so anything heading to production still needs a human review
- Feeding a live product into the tool means sending screens and behaviour to servers in India and/or the United States, with consent to that transfer implied by use of the service and EU residency available only on Enterprise
- Pricing documentation contradicts itself: an obsolete grid in the documentation FAQ, two credit scales on the pricing page and a roll-over rule that differs from the terms. Budget on the pricing page's figures and get the credit rules confirmed in writing
- Liability is capped at the higher of the fees paid over 12 months or USD 100, which is low against the cost of a design decision gone wrong
- Dispute resolution is committing: mandatory arbitration in Delaware for US users, exclusive jurisdiction of the courts of Bengaluru for everyone else
- Two contractual details are easy to miss: the terms forbid using the output to compete directly with Figr, and health or payment data must not be submitted without an explicit agreement, while the minimum age of 18 in the terms sits oddly next to a privacy policy that only addresses children under 13
Setup & Integrations
Technical difficulty
Low for the everyday user. Figr quotes about 20 minutes to get running, up to an hour on a complex product: you supply four or five screenshots, Figma tokens, CSS files or a documented design system, install the Chrome extension for live capture, and let Figr auto-explore the flows and set up the design system from its backend. A Figma plugin handles import and export, and no design skill is required. A 15-minute onboarding call is offered. Heavier work is Enterprise-side: SSO, RBAC and audit logs need IT involvement, and the MCP server must be registered in the IDE.
Deployment
Integrations
Behind Figr
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Figr.
Frequently asked questions
What is Figr?
How is it different from other AI design tools?
Is there a free plan, and what does the cheapest paid plan cost?
How do credits work?
Is my data used to train the AI models?
Where is my data hosted?
What security certifications does Figr hold?
Can I use Figr with my existing Figma files?
Is there an API or a mobile app?
What happens to my data if I delete my account?
Should you pick Figr?
Figr knows where it stands and says so: it works upstream of the design canvas and hands the result to Figma instead of trying to replace it. That restraint is the product's strength. Where most AI design tools take a prompt and return a screen, Figr spends its effort on context, capturing the running product, reading the design system, absorbing PRDs, research and analytics, then showing its UX reasoning before it draws. For a team whose product already carries years of behaviour, that order matters more than raw generation speed.
The privacy posture is written down rather than implied: no training on customer data in both the terms and the enterprise terms, SOC 2 Type II audited annually, encryption at rest and in transit, and a named list of subprocessors. The free plan is permanent, and read-only guests cost nothing on any tier.
The reservations are real, and more editorial and contractual than technical. GDPR compliance is never claimed in so many words, there is no DPO and no Article 27 representative, and EU data residency is an Enterprise-only feature while everything else sits in the United States or India. The pricing documentation contradicts itself in three places, and an internal editorial note was left published on the pricing page. Liability is capped low and disputes go to arbitration in Delaware or to the courts of Bengaluru.
The company is young: domain registered in 2023, seed funding behind it, named customers in front of it. If you run an existing product, the free plan costs nothing to test. If you operate in the EU or in a regulated sector, settle data residency, the DPA and the credit rules in writing first.
- Choosing a selection results in a full page refresh.
- Opens in a new window.