Bolt.new for Product Managers: Plan Mode, Prototypes and Slides

Bolt.new (Bolt) suits a product manager who wants to test an idea as a working app instead of a written spec. You describe the feature, Bolt's agent writes a JavaScript web app with a real database and sign-in, and stakeholders get a live link to click through. The core product-manager kit of Plan Mode, comments, sharing roles and Bolt Slides runs on every plan, including Free at $0 a month, which carries 1M tokens a month with a 300K daily limit ^[1]. The paid gates that matter are narrow. The Max agent and private published sites start on Pro, and adding your own design system needs paid Teams. Our verdict: a strong tool for validation and pitching, with the handoff to engineering running through GitHub.

Scope the idea in Plan Mode before Bolt writes code

Plan Mode lets you plan and ask questions without changing any code, with the project's codebase and recent messages as context, and it can pull real-time information from trusted web sources, listing them at the top of the reply ^[2]. For discovery work that is the right place to pressure-test scope. Ask which screens the idea needs, what data it has to store, and which edge cases a first version can skip. When the plan reads right, the Implement this plan button switches to Build Mode. A new project can start in Plan Mode straight from the homepage; for a mobile app, say so in the first prompt.

Plan Mode is cheaper, not free. Bolt says it uses fewer tokens than Build Mode, but never zero. Tokens measure the work Bolt does reading prompts and project files, so a bigger project burns more per message as it grows. The plan selector's figure of about 100 messages for 10M tokens is Bolt's estimate, not an allowance. One habit pays off quickly: group related questions into a single message, because Bolt itself notes that many small prompts about one screen cost more than one clustered prompt.

Bolt offers two agents and picks the model behind each one. Standard runs on every plan and is the only agent on Free; Bolt pitches it for small and medium apps, interface updates and well-defined tasks. Max adds deeper reasoning for large apps, interconnected features, refactoring and open-ended work, spends more tokens and needs a paid plan. Most early prototypes sit comfortably in Standard's lane. Max earns its cost when one change ripples across many screens, or when you build against a design system, where Bolt recommends it.

A prototype that stores data and survives a usability test

Bolt says the prototype it builds is a working app with a real database and real sign-in accounts, and that the prototype and the finished product are the same app with the same code. The backend is Bolt Database, built on PostgreSQL and running on Supabase infrastructure managed by Bolt; it comes with every plan and appears automatically when the app needs one, such as for a login form. That changes what a test session can reveal. Testers sign up, save records and return to them, so you observe real behavior instead of guessing from a static screen.

Iterating has its own rules. Bolt saves a version after every processed prompt, and restoring a version costs no tokens, so roll back a failed experiment rather than asking the agent to undo it. Restoring never rolls back the database, which means test records from an abandoned variant stay behind. Copy and styling tweaks go through visual edits with the Select button, free while you work and charged only when the batch is saved. Code view edits use no tokens at all. That is handy when a label just needs one word changed.

Publishing puts the prototype on a free bolt.host address, and the Publish and Update buttons spend no tokens. On Free every published site is public, even though Free projects themselves can be private. Pro adds private sites shared with invited emails, the safer choice for an unannounced feature; opening a private site to a whole trusted email domain takes Teams. Hosting limits apply to the whole account. On Free, every published site shares 10 GB of bandwidth and 333,333 requests a month, and sites stop serving once either runs out.

Collecting feedback without booking another meeting

Sharing works on every plan through three roles. Viewers can open, duplicate, view code and comment, but cannot prompt or edit. Editors can prompt, edit code, view environment variables and publish. Co-owners hold full control and must belong to your own team, so people outside a team or on a personal plan get Viewer or Editor only. One catch is easy to miss. A project that depends on environment variables may not fully work for Viewers, which matters when an executive opens the link.

Preview comments arrived in October 2026 on all plans. Anyone with a role can leave and answer comments pinned to the project preview, though not on a published site. Only Editors and Co-owners can send comments to Bolt as a prompt, and that spends the sender's own tokens rather than the owner's. Sending several comments together usually costs fewer tokens than sending them one by one. The tidy routine follows from that: designers and engineers annotate, then the product manager batches the accepted points into one send.

Real-time multiplayer, added in April 2026, lets several people work inside one project at once. There is still a single shared chat thread, and Bolt processes one prompt at a time while queuing the rest, with each prompter paying from their own account. Inside a Teams workspace, inviting outsiders such as a customer advisory group stays off until an admin enables external collaboration. Team members also get the Viewer role on team projects by default.

Turning the review into a Bolt Slides deck

Bolt Slides launched in July 2026 on every plan, Free included, and builds a deck from a prompt with a title slide, content slides, speaker notes and a speaker view. For a product review the useful part is that a deck can embed working prototypes and live data, and you can feed it a doc, meeting notes, a website or an existing deck as input ^[3]. Prompting adds interactive elements such as sliders, clickable maps and collaborative idea boards. Bolt argues that a web deck living at one URL keeps every viewer on the same version.

Publishing a deck creates a link that anyone can open without a Bolt account. Viewers cannot edit and never see speaker notes, and restricting access means prompting Bolt to add a password screen. Watch the PDF export. It flattens interactive content into a static image, so an embedded prototype survives only as a picture. AI-generated images in slides are paid-only. Decks created before the September 2026 Slides update lack reordering and speaker view, so start a fresh Slides project to get those controls.

Keeping prototypes on brand with Figma, Stitch and design systems

Designs arrive by two routes. Bolt uses Anima to convert a single Figma frame, not a whole file, into an app, and since January 2026 a frame can also land in a project already under way. Google Stitch hands a design over one way, with screenshots, HTML and a prefilled prompt, and you need a Bolt account on the receiving end. Our Figma and Stitch guide walks through both imports step by step.

Design systems split by plan. Everyone can build with the pre-loaded Chakra, Material UI and Shadcn systems, but adding your own design system requires a paid Teams plan. Bolt accepts files, public websites, a public GitHub repo, a Storybook or NPM packages as sources, and says repos and packages give the best results. Each team gets 10 adds or syncs per week, and every sync keeps a version history. A team design system can also be selected for a Slides deck. Pro users may upload brand guidelines into a project, which helps, but that is not a design system.

Budgeting for a product team, and the results Bolt claims

Free and Pro are personal accounts for one named individual. Pro starts at 10M tokens a month for $25 per month billed monthly or $18 per month billed yearly ($216 a year). Teams costs $30 per member per month billed monthly or $27 per member per month billed yearly ($324 a year per member), and each paid member gets a separate allotment that is never pooled. By our arithmetic, three members at the 10M level pay $90 per month billed monthly or $81 per month billed yearly ($972 a year). Rollover and upgrade rules sit in our pricing guide.

The deciding factor is collaboration, not tokens. A solo product manager gets everything above except a custom design system on Pro. Once designers and engineers need prototypes built from the company's component library, Teams becomes the plan; by our arithmetic each member costs $5 more per month than one Pro account at the 10M level when billed monthly, or $9 more per month when billed yearly. Free members of a team can view projects. They cannot prompt.

Bolt's product-teams page, as we read it in November 2026, says prototypes are 7x more likely to be prioritized, 200% faster to ship and 10x faster to stakeholder buy-in than static PRDs ^[4]. Bolt does not explain how these figures were measured. The customer accounts are more concrete. Sai Kishore, Head of Product at Zoho, says Bolt cut the design-to-dev cycle nearly in half and let the team test 10 to 15 feature variations in the time it used to take to finalize two. Julien Menard, CTO of Digital Virgo, reports cutting time to launch from 12 months to a few months, with Bolt quoting the result as 3.5 months in one post and three to four months in another.

The bottom line

Bolt fits product managers whose real bottleneck is getting a believable version of an idea in front of people. Plan Mode, comments and Slides cost nothing to try, Pro unlocks the Max agent and private sites, and Teams is where company-wide design systems and per-member allotments live. Keep an eye on token burn as a prototype grows, and remember that every Free site is public and that a version restore leaves the data untouched. Judge it on one real feature first.