Hero image for v0 vs Lovable vs Bolt vs Figma Make: Which AI App Builder Can an Agent Read? (2026)

v0 vs Lovable vs Bolt vs Figma Make: Which AI App Builder Can an Agent Read? (2026)

comparisondesign-to-codeai-app-builders

Why agent-readability is the real question

For an agent, the right builder is the one that exposes the real repository, design context, and a verification loop. As of 2026-10-24, v0 and Lovable document Git-based code access on their own pricing and docs pages, which is the main reason they score highest on code exit in this review: see the v0 GitHub docs and Lovable subscription plans.

Four browser-based builders come up in most shortlists: v0 by Vercel, Lovable, Bolt, and Figma Make. Each generates app code from a prompt, but they differ on whether that code lives where an external agent can reach it. This guide asks one question: which tool exposes code, design context, and verification surfaces an agent can read and act on programmatically? For a wider set, see our agent-readable app builders we compared previously.

This review is based on official documentation, pricing pages, and community reports — we did not run the tool hands-on. Treat every score below as a reading of published claims, not a test result.

What “agent-readable” means

We score each tool from 0 to 5 on five dimensions: code exit, design-system input, verifiable output, cost predictability, and connectors/MCP. The rubric is this post’s own editorial definition, and the scores are the author’s assessment from pages such as the v0 pricing page, not measurements.

  • Code exit: can an agent read and write the real repository, through Git sync, a branch, or a download?
  • Design-system input: how design context such as design systems, component kits, frames, and documents reaches the model.
  • Verifiable output: whether changes appear in a checkable form, such as a pull request, a CI check, or a preview.
  • Cost predictability: whether an agent loop’s cost can be estimated from published rules before it runs.
  • Connectors/MCP: whether external tools can be attached to the builder.

A score of 0 means no support is visible in the cited pages. A score of 5 means the strongest documented support.

Comparison table

v0 scores highest on code exit and verifiable output, Figma Make on design-system input, and Lovable on connectors. Figma Make scores lowest on cost predictability because its pricing was not verified, and Bolt is next lowest.

Tool Code exit Design-system input Verifiable output Cost predictability Connectors/MCP
v0 by Vercel 5: GitHub sync on Free; working-branch pushes 2: Design Mode edits; no design-system input documented in our sources 5: PRs and visible CI checks 3: 7 messages/day on Free; per-model token rates 2: MCP docs exist; capabilities not assessed
Lovable 4: Git sync on all plans; editing and download not on Free 3: Design systems on Pro+; design templates on Business+ 3: Preview and publish; no PR/CI flow documented 3: Per-prompt credit examples; 1.70 vs 2.00 dispute 4: Custom connectors on all plans
Bolt 0: Download and GitHub export not verified 3: Design System Knowledge and NPM registries on Teams 2: Publish and hosting; no PR/CI flow documented 1: Token use scales with project file-system size 3: Connectors and MCPs on Pro and above
Figma Make 3: Code-backed prototypes; GitHub push and Git articles listed, not read 5: Make kits, npm packages, Figma frames, PDFs 2: Prototype preview only; no PR/CI documented 0: Pricing not verified 3: MCP connectors listed on the page; scope not read

v0 by Vercel: the strongest PR and CI surface

v0 gives the strongest agent loop here. GitHub sync is on the Free tier, it creates a private repository on connect, and it pushes to a working branch with pull requests and visible CI checks, per the v0 GitHub docs, as of 2026-10-24.

The GitHub flow

  • On connect, v0 creates a private repository and pushes the current code. From then on, GitHub is the source of truth.
  • Chats start from a base branch, but v0 pushes changes to a working branch rather than the base.
  • Publishing goes through a pull request, and CI checks appear on it. Pull Changes and Fix Conflicts handle sync.
  • Deployment permissions follow the Vercel role, not the v0 role.
  • Existing repositories can be imported from GitHub.

For an agent, a change becomes a branch, a diff, and a check. That is the clearest verification path in this set. Our agent-verifiable output in design-to-code tools explains why that matters.

Pricing and token rates

Free is $0, with a one-time $5 credit for new users and a limit of 7 messages per day. Plus lists at $30 per user per month, with $30 of included monthly credits per user; the page also shows a struck-through $90 alongside it. Business is $100 per user per month, with training opt-out on by default, per the v0 pricing page, as of 2026-10-24.

Model token rates, per 1M tokens (input / output):

  • v0 Mini: $0.20 / $1.20
  • v0 Pro: $2 / $10
  • v0 Max: $5 / $25
  • v0 Max Fast: $10 / $50

Design Mode provides visual editing. v0 documents MCP integrations at v0 MCP docs. We link that page and make no claims about what it supports. Cost predictability scores 3 because the daily message cap is clear, but the pages we reviewed do not publish cost per action.

Lovable: Git sync on every plan, credits priced per prompt, one example disputed

Lovable is among the most agent-accessible builders on code exit, because Git sync to GitHub, GitLab, and Bitbucket is included on every plan, including Free. Editing and downloading code are not on Free, per its subscription plans docs, as of 2026-10-24.

Plans and code access

Free is $0, with 5 daily build credits capped at 30 per month, a monthly Cloud grant of 20, and a monthly AI grant of 4. Pro starts at $25 per month with 100 monthly credits, or $21 per month billed annually. Business starts at $50 per month, or $42 per month billed annually. Enterprise is custom. Figures are from the Lovable pricing page, as of 2026-10-24.

Downloads are available on Pro, Business, and Enterprise. Custom domains and design systems require Pro or higher, and design templates require Business or higher. Custom connectors work on all plans, Free included. The pricing page says you own your code.

Credits per prompt

Plan mode costs 1 credit per message. The pricing page gives these build-mode examples:

  • “Make the button gray”: 0.50 credits
  • “Remove the footer”: 0.90 credits
  • “Add authentication”: 1.20 credits
  • “Build me a landing page, use images”: 1.70 credits

The Lovable sources disagree on the last example. A third-party guide says the docs list it at 2.00 credits. We could not resolve this from the pages, so we report both values and do not pick one.

Lovable’s AI docs describe a built-in AI connector for deployed apps, with per-workspace rate limits that return 429 and 402 errors. Cost predictability scores 3: the published examples make estimation possible, but the 1.70 versus 2.00 dispute limits confidence.

Bolt: tokens, project-size costs, and gated connectors

Bolt has the least predictable cost model here after Figma Make’s unverified pricing. Its FAQ says most token usage comes from syncing the project’s file system to the AI, and connectors and MCPs are excluded from Free, per the Bolt FAQ and Bolt pricing, as of 2026-10-24.

The token model

The FAQ states it directly: “most token usage is related to syncing your project’s file system to the AI: the larger the project, the more tokens used per message.” For an agent running many turns, cost grows with the codebase, not only with the task.

Free is $0, with 1M tokens per month, a 300K daily limit, and bolt.host hosting with 10 GB of bandwidth. It has no custom domains, shows a Bolt badge on published sites, and does not roll over unused tokens. Pro is quoted here only at its annual rate: $18 per month billed annually ($216 per year), with 10M tokens per month, rollover, custom domains, and connectors and MCPs. Teams is $27 per member per month billed annually ($324 per year), adding Design System Knowledge, private NPM registries, and SSO. Figures are from the Bolt pricing page, as of 2026-10-24.

Connectors and code exit

Connectors and MCPs are excluded from Free and included on Pro, Teams, and Enterprise. Bolt uses WebContainers from StackBlitz, per its FAQ. We did not find Bolt code download or GitHub export verified in the documentation we reviewed, so we score code exit at 0 and make no claim that it exists.

Forge and billing dates

Bolt Forge was a research preview from Sep 14 to Oct 14, 2026, with open-source models including GLM 5.3, GLM 5.3 Flash, Kimi K3, and DeepSeek v4 Pro. Per the Bolt release notes, the preview has ended. Billing moves to Link (Stripe) from Nov 1, 2026. Subscriptions are non-refundable, per the FAQ.

Figma Make: design context first, code exit still thin

Figma Make leads on design-system input, since make kits, npm packages, Figma frames, and PDFs feed the model, per its Figma Make page. Code exit is thinner: prototypes are code-backed and editable, but local-codebase support is labelled Beta and “coming soon” on that page.

Design context

The page says, “Whatever you build uses your design context, from systems to images.” Make kits and npm packages bring design systems in. Figma frames and PDFs can be attached as context. Files can also be copied to the canvas for tweaking. This makes Figma Make the strongest input layer in the set, even though its code exit scores below v0 and Lovable.

Code exit and backend

The page says everything you build is “code-backed and visually editable, so you can easily jump between code and canvas.” A help center article covers editing the code of a functional prototype or web app. Figma Make also has a built-in CMS through Payload, described in a help article about adding a backend.

The help center also lists “Push from Figma Make to GitHub” and “Use Git with Figma Make and code layers.” We did not read those articles, so we score code exit 3 and do not claim the GitHub path works.

The local-codebase feature is the gap. The marketing page labels “Make on your local codebase” as Beta and says “Coming soon.” The help center has a “Make in your local codebase” article, but its status is unclear from the pages we read. Treat local-repository support as announced, not available. Check the Figma Make help center before you plan a workflow around it.

Connectors and pricing

The page also lists MCP connectors, with a help article on verified partner connectors. We did not read that article, so the connectors score of 3 is low confidence. Figma Make pricing is not verified in this review. Check Figma’s plans page before you budget, and read the 0 on cost predictability as a gap in our evidence, not a verdict on the price.

Head-to-head: where each tool wins for agents

v0 wins on code exit and verifiable output, Figma Make on design-system input, and Lovable on connectors. On cost predictability, Lovable and v0 tie at 3, and both sit ahead of Bolt (1). No single tool wins every dimension.

  • Code exit: v0 (5) over Lovable (4), then Figma Make (3). v0’s GitHub sync is on Free, while Lovable restricts editing and downloads on Free.
  • Design-system input: Figma Make (5). Lovable and Bolt (3 each) gate design features by plan tier.
  • Verifiable output: v0 (5). Pull requests and CI checks give an agent a diff and a pass-or-fail signal.
  • Cost predictability: Lovable and v0 (3 each) over Bolt (1). Lovable publishes per-prompt examples, while Bolt’s token use grows with project size. Figma Make (0) is unscored on price because its pricing was not verified.
  • Connectors/MCP: Lovable (4), with custom connectors on all plans. Bolt and Figma Make (3 each) list connectors or MCP support, but Bolt gates them to Pro and above.

FAQ

These four questions come up most often for agent workflows. Short answers, as of 2026-10-24: Lovable syncs code on Free but does not allow download, v0 syncs on Free, Bolt Forge has ended per its release notes, and Figma Make’s local-codebase support is announced but not yet available.

Can I get my code out of the free tier?

On Lovable, Git sync works on Free, but editing and downloading code do not, per its subscription plans docs. On v0, GitHub sync is on Free, so code reaches your repository. For Bolt, we could not verify code download or GitHub export in the documentation we reviewed, so we would not plan on it.

Is Bolt Forge available?

No. Bolt Forge was a research preview from Sep 14 to Oct 14, 2026, and its release notes say it has ended. Do not plan around its open-source models, which included GLM 5.3 and Kimi K3. Billing moves to Link (Stripe) from Nov 1, 2026, per the Bolt release notes.

Which builder has the most predictable costs for agent loops?

Lovable and v0 tie at 3 on the published evidence, both ahead of Bolt (1). Lovable publishes per-prompt credit examples from 0.50 to 1.70, and plan mode costs 1 credit per message. Bolt warns that larger projects consume more tokens per message. Budget with margin, because Lovable’s landing-page example is disputed at 1.70 versus 2.00.

Does Figma Make work in a local repo?

Not as far as we can verify. The marketing page calls local use “coming soon” and labels it Beta. The help center has a “Make in your local codebase” article, but its status is unclear. Treat local-repository support as announced rather than available, and check Figma’s help center before you plan a workflow around it.

The Bottom Line

Choose v0 for agents that live in Git and CI, Lovable for Git sync at every price tier, Bolt only if size-dependent token costs fit your budget and you pay for Pro connectors, and Figma Make for design-context-heavy work with a thinner code exit. Prices are as of 2026-10-24.

  • v0: the best fit when the agent’s loop runs through GitHub branches, pull requests, and CI checks.
  • Lovable: Git sync at any tier, with published credit examples. Editing and downloads need a paid plan.
  • Bolt: a fit only if you accept token costs that grow with project size and you pay for Pro to get connectors.
  • Figma Make: a fit for design-heavy work. Re-check its code exit and pricing before you commit, since neither is verified here.

How This Guide Was Built

This guide is based on official pricing pages, documentation, and community reports as of 2026-10-24, including the Lovable pricing page. We did not test any tool hands-on. The five-dimension rubric is defined in this post, and we flag source disagreements, including Lovable’s 1.70 versus 2.00 landing-page credit example.

  • Primary sources: Lovable pricing and subscription plans docs, Bolt pricing, and v0 pricing and GitHub docs.
  • Supporting sources: Lovable AI docs, Bolt FAQ and release notes, and Figma Make pages.
  • Scores: the author’s assessment from those pages, not measurements.
  • Unverified: Bolt code download and GitHub export, Figma Make pricing and plan limits, v0 MCP capabilities, the Figma help-center GitHub and Git articles, the Figma verified-partner connectors article, and whether Figma’s local-codebase feature has shipped. We treat these as unverified rather than scoring them as present.
  • Currency: pricing changes. Check the linked pages before you buy.