# Bolt pricing: plans, tokens, and why big projects cost more

Bolt charges tokens, and most of them go on reading your project, not your prompt. The plans, what each tier buys, and how to make a month's tokens last.

Published 2026-10-07 by Richard Kaminsky and Mitchell Lipyansky, the co-founders of Poly. Canonical: https://usepoly.co/bolt-pricing
Poly is a multiplayer AI coding workspace: a shared room where your team works with one AI agent, together. Free to start: https://usepoly.co/

**Bolt has a free plan with 300,000 tokens a day and 1 million a month; Pro from $25 a month for 10 million tokens, with larger tiers at $50 for 26 million and $100 for 55 million; Teams at $30 per member, where each member gets their own tokens; and custom Enterprise plans. Yearly billing cuts the $25 Pro to $18 a month. Paid tokens roll over for one extra month. Bolt's own pricing page explains the catch: most tokens go on "syncing your project's file system to the AI," so the larger the project, the more each message costs. In a shared project, the person who sends a prompt pays for it.**

## Key takeaways

- **Plans:** Free; Pro from $25 a month ($18 billed yearly); Teams $30 a member ($27 yearly); Enterprise custom.
- **Free:** 300,000 tokens a day, 1 million a month, the Standard agent only, and a Made in Bolt badge on sites.
- **Pro tiers:** 10 million tokens for $25, 26 million for $50 and 55 million for $100, with Pro 200 above.
- **Tokens last two months** on paid plans; tokens bought as a reload never expire.
- **Teams don't pool tokens:** each paid member has their own, and in a shared project the prompter pays.

*Bolt's plans, from bolt.new/pricing and Bolt's help center, October 7, 2026.*

|  | Free | Pro | Teams | Enterprise |
| --- | --- | --- | --- | --- |
| Price | $0 | From $25 a month ($18 yearly) | $30 a member ($27 yearly) | Custom |
| Tokens | 300K a day, 1M a month | From 10M a month | From 10M a member | Custom |
| Unused tokens roll over | No | Yes, two months | Yes, per member | Custom |
| Agents | Standard | Standard and Max | Standard and Max | Standard and Max |
| Hosting | 10 GB, 333K requests, hard cap | 30 GB, 1M requests, pay as you go | Same as Pro | Custom |
| Custom domain, no badge | No | Yes | Yes | Yes |
| Single sign-on, audit logs | No | No | No | Yes |

## How much does Bolt cost?

From Bolt's pricing page and help center ([Bolt](https://bolt.new/pricing), [billing](https://support.bolt.new/account-and-subscription/billing)):

- **Free, $0:** 300,000 tokens a day and 1 million a month, public and private projects, website hosting with a Made in Bolt badge, 10MB uploads and unlimited databases.
- **Pro, from $25 a month:** "Start at 10M tokens per month," no daily limit, no badge, custom domains, private site sharing, 100MB uploads, more database capacity, AI image editing and unused tokens that roll over.
- **Teams, $30 per member a month:** everything in Pro plus central billing, team access management, admin controls and user provisioning, private npm registries and design-system knowledge.
- **Enterprise, custom:** single sign-on, audit logs, compliance support, data retention policies, 24/7 support and service agreements.

Bolt says yearly billing saves "up to 28%": the $25 Pro becomes $216 a year, or $18 a month, and Teams $27 a member. The larger tiers save less; Bolt's billing guide says "yearly pricing at 10% off," and its example prices the 26 million token plan at $540 a year.

## How many tokens does each Bolt plan include?

- **Free:** 300,000 a day, 1 million a month. Unused daily tokens don't carry over, the monthly allowance resets on the 1st, and you can still edit code by hand in Code view for free.
- **Pro (10M):** 10 million a month for $25.
- **Pro 50:** 26 million a month for $50.
- **Pro 100:** 55 million a month for $100.
- **Pro 200 and up:** larger allowances, listed in the plan picker.

Bolt doesn't say how many tokens a message uses, because it depends on the project. "Most of your token usage comes from Bolt reading, understanding, and syncing your project files, so larger projects use more tokens per message" ([Bolt](https://support.bolt.new/account-and-subscription/tokens)).

## Do Bolt tokens roll over?

Yes, on paid plans. Since July 1, 2025, unused tokens "stay valid for two months from the start of the billing cycle in which you received them." The oldest tokens are spent first. Rolled-over tokens need an active paid plan: cancel, and you lose them at the end of the billing cycle, though resubscribing before they expire brings them back. Free tokens don't roll over ([Bolt](https://support.bolt.new/account-and-subscription/tokens)).

## Can you buy more Bolt tokens?

Bolt calls upgrading to a higher tier "the most cost-effective option." Reloads, a one-off top-up, are only for people on the highest monthly Pro plan or any yearly Pro plan, and "Any tokens purchased as a reload do not expire." Bolt doesn't publish a reload price: it "varies based on your plan" and appears when you go to buy, with higher yearly tiers paying less per 10 million tokens ([Bolt](https://support.bolt.new/account-and-subscription/tokens)).

## What do the Standard and Max agents cost?

Bolt replaced its model picker in May 2026 with two agents, and "Bolt handles model selection behind the scenes" ([Bolt](https://support.bolt.new/building/using-bolt/agents)):

- **Standard:** fast and token-efficient, the default for most work. "Only Standard is available on the free plan."
- **Max:** deeper reasoning for large codebases and hard problems, on paid plans. It "uses more tokens," so Bolt suggests keeping it for complex tasks.

Both draw on the same token balance.

## What are Bolt Forge and Bolt Lite?

Bolt Forge, launched September 14, 2026, is a third agent that runs open-source models (GLM 5.3, GLM 5.3 Flash, Kimi K3 and DeepSeek v4 Pro) with "up to 50x more" usage than Standard, measured on GLM 5.3 Flash. It's a research preview until October 14, 2026, free on individual Pro plans until then. After the preview it stays only on Bolt Lite, a $9 a month plan with Forge only, which new and free users join from a waitlist. Prompts and code sent through Forge are anonymized and shared with Bolt's partners to train open-source models, and Forge isn't on Teams or Enterprise ([Bolt](https://support.bolt.new/account-and-subscription/bolt-forge)).

## Does Bolt include hosting?

Yes. Bolt Cloud hosting comes with every plan ([Bolt](https://support.bolt.new/cloud/hosting/plans)):

- **Free:** up to 10 GB of bandwidth and 333,333 requests a month, about 10,000 to 15,000 visits by Bolt's estimate. Sites go offline at the limit until the next cycle.
- **Paid plans:** up to 30 GB and 1 million requests, often 100,000 or more visits, plus pay-as-you-go traffic up to a spending cap you set. Bolt doesn't publish the pay-as-you-go rate.

The limits cover your whole account, not each site. Custom domains need a paid plan. Bolt builds iOS and Android apps through Expo at no extra Bolt charge, but publishing them needs paid Apple and Google developer accounts ([Bolt](https://support.bolt.new/integrations/expo)).

## How do you use fewer Bolt tokens?

From Bolt's guide to token efficiency ([Bolt](https://support.bolt.new/best-practices/maximizing-token-efficiency)):

- **Plan first,** in Plan mode, which "uses fewer tokens" because it doesn't change code.
- **Use Standard** unless a task needs Max.
- **Name the files or functions** to change rather than the whole app.
- **Use buttons instead of prompts:** publishing and Version History cost no tokens.
- **Don't click Attempt fix over and over;** each try uses tokens.
- **Turn connectors on only when needed,** clear context, and keep the project small.

## Who pays when a team shares a Bolt project?

The person who prompts. "Tokens are not shared among team members" on Teams, and rolled-over tokens "stay linked to the person who accrued them." Bolt's multiplayer mode, added in April 2026, works once the project is set to public: everyone shares one chat, Bolt runs one prompt at a time and queues the rest, and "Bolt uses tokens from the prompter's account rather than from the project owner's." Only the owner can connect GitHub or MCP servers ([Bolt](https://support.bolt.new/building/using-bolt/collaborate)).

## Does Bolt train on your projects?

From October 7, 2026, by default. A new Model training setting lets Bolt use your projects and chats to train AI models; it's on unless your account is based in the EU, UK or Switzerland. You can turn it off in account settings at any time without changing your plan, and it covers content created or sent to Bolt's AI on or after October 7 ([Bolt](https://support.bolt.new/settings/account-settings)).

## Does Bolt have a student discount?

Bolt publishes no student or regional discount as of October 7, 2026. A friend's referral link gives 20% off a paid plan ([Bolt](https://support.bolt.new/release-notes)).

## How has Bolt's pricing changed?

- **July 2025:** paid tokens started rolling over for one extra month.
- **April 2026:** the original v1 agent closed to new projects (April 13), and multiplayer launched.
- **May 2026:** the Standard and Max agents replaced the model picker.
- **July 2026:** referrals pay cash, and the friend gets 20% off.
- **August 3, 2026:** the v1 agent retired.
- **September 14, 2026:** Bolt Forge preview and the $9 Bolt Lite.
- **October 7, 2026:** model training on by default outside the EU, UK and Switzerland.

Bolt's release notes list each change ([Bolt](https://support.bolt.new/release-notes)).

## Bolt vs Lovable, v0, Replit and Base44 pricing

- **Lovable:** Pro $25 for 100 credits plus 5 a day; collaborators spend the owner's workspace credits ([Lovable pricing](/lovable-pricing), [Lovable vs Bolt](/lovable-vs-bolt)).
- **v0:** Plus $30 a user; free with 7 messages a day ([v0 pricing](/v0-pricing)).
- **Replit:** Core $20 with $20 of credits ([Replit pricing](/replit-pricing)).
- **Base44:** Starter $20 ([Base44 vs Lovable](/base44-vs-lovable)).

Bolt is the one that charges whoever prompts, which suits a group where everyone pays their own way and hurts one where a single person funds the project.

## Building as a team

Bolt's multiplayer asks for a public project and runs prompts one at a time. Poly (usepoly.co) is built for a team building one app with one AI: a shared browser room rather than a public project, where everyone sees each step live and any member can approve a change before it lands. It runs Claude Code and Codex. Free to start. [What is Poly?](/what-is-poly)

## Common questions

**How much does Bolt cost?**

Bolt is free with 300,000 tokens a day and 1 million a month. Pro starts at $25 a month for 10 million tokens, or $18 a month billed yearly, with larger tiers at $50 and $100. Teams costs $30 per member, and Enterprise is custom.

**How many tokens does a Bolt message use?**

It depends on the project. Most tokens go on Bolt reading and syncing your project files, so the same prompt costs more in a larger project. The Max agent and repeated automatic fix attempts use more too.

**Do Bolt tokens roll over?**

On paid plans, yes: unused tokens stay valid for two months from the start of the billing cycle they were issued in, and need an active paid plan. Free tokens don't roll over. Tokens bought as a reload never expire.

**Do Bolt Teams members share tokens?**

No. Each paid member gets their own monthly tokens, and in a shared project the person who sends a prompt pays for it, not the project owner.

**What is Bolt Lite?**

A $9 a month plan, joined from a waitlist, that runs only Bolt Forge, Bolt's agent built on open-source models such as GLM 5.3 and Kimi K3. Forge is free on individual Pro plans during its preview, which ends October 14, 2026.

More guides: https://usepoly.co/guides · Security: https://usepoly.co/security · Pricing: https://usepoly.co/pricing