01
Overview
Bolt Forge is available as a research-preview agent inside Bolt.new, but using it is different from simply choosing another premium model. Forge has its own eligibility rules, opt-in flow, model lineup, usage meter and privacy tradeoff.
This guide covers the practical workflow for getting into Forge and using it without confusing the temporary launch promotion with Bolt’s normal Standard or Max agents.
02
Who can access Bolt Forge?
At launch on September 14, 2026, Forge is available to individual Bolt Pro users.
Bolt also announced Bolt Lite as a lower-cost access route, but Lite is being distributed in waves rather than opened to everyone immediately. Bolt says the first Lite seats are expected by September 21, 2026.
Teams and Enterprise workspaces are excluded from Forge at launch.
If Forge does not appear in your workspace, the first thing to check is whether you are on an eligible individual account rather than a Teams or Enterprise workspace.
03
Step 1: Open the agent picker
Inside an eligible Bolt.new project, open the agent picker.
Forge appears as a third agent alongside Standard and Max. Bolt describes Forge as a separate agent rather than a single model because it can route work through several open-weight models.
If you are working on an important existing project, duplicate the project before switching. Bolt itself recommends doing this because Forge’s model lineup is experimental.
04
Step 2: Choose Forge and review the opt-in
When you switch into Forge, Bolt displays a consent screen.
This is not a one-time account-wide setting hidden in general terms. Bolt says the consent screen appears each time you enter Forge so the data-sharing tradeoff remains visible.
If you decline, you stay with Standard or Max and your work is not moved into the Forge training-data workflow.
If you accept, Forge sessions may be anonymized and used with Arcee AI to help train open-weight models.
That means the opt-in is part of using Forge, not a separate optional feature after you are already inside it.
05
Step 3: Start with GLM 5.3 Flash
Bolt lists GLM 5.3 Flash as the default Forge model at launch.
For most users, starting with the default is sensible because Bolt specifically notes that some experimental alternatives consume the Forge allowance faster.
The launch model lineup is:
- GLM 5.3 Flash — default;
- GLM 5.3;
- Kimi K3 — experimental;
- DeepSeek v4 Pro — experimental.
The lineup can change during the research preview, so always check the live agent picker before assuming a model is still available.
06
Step 4: Build as normal inside Bolt
Once Forge is active, the rest of the experience remains inside Bolt’s existing app-building environment.
You can prompt the agent to create or modify an app, website or prototype. Bolt says Forge projects still run through WebContainers in the browser, which is part of the cost structure that allows the larger experimental allocation.
Forge is especially well suited to rough exploration: trying an idea, rebuilding a section, testing several approaches and iterating without consuming the same premium allocation used by Standard or Max.
07
Step 5: Watch the Forge usage bar
Forge uses its own monthly usage bar.
During the preview, Bolt says individual Pro users get up to 50X more Forge usage at no extra cost through October 14, 2026. The bar resets on the user’s normal renewal date.
There are no daily limits inside that Forge allocation. When it reaches 100%, Bolt says the system switches the user back to Standard rather than charging an automatic Forge overage.
Forge usage is separate from Standard and Max usage, so experimenting in Forge does not mean the same usage pool is being consumed.
08
How model choice affects usage
Not every Forge model consumes the allowance at the same pace.
Bolt specifically says Kimi K3 and DeepSeek v4 Pro burn through the Forge allocation faster than the GLM options. The company does not publish a universal per-prompt conversion that lets users calculate an exact cost before every request.
Because of that, treat the usage bar as the source of truth for your own account rather than trying to infer a fixed token-to-Forge conversion.
09
When should you use Standard or Max instead?
Bolt recommends keeping complex production work in Standard or Max.
Forge is experimental. Its strongest use cases are brainstorming, MVPs, prototypes and iterative builds where the extra usage room is valuable.
Switch back to Standard or Max when:
- the project is production-critical;
- reliability matters more than the larger experimental allowance;
- you do not want the current Forge session to enter the opt-in training workflow;
- you need functionality that Forge does not yet support.
Switching back stops new Forge data sharing for future work, although data already used for model training cannot be removed from a model after the fact.
10
Can Bolt Forge upload PDFs?
Not at launch. Bolt explicitly says Forge cannot take PDF uploads yet.
If your workflow depends on PDF context, use another supported Bolt path until Forge adds that capability.
11
Why Forge may not be showing
There are several launch-period reasons Forge may be missing:
- You are not on an eligible individual Pro account.
- You are working in a Teams or Enterprise workspace, which is excluded at launch.
- You are waiting for Bolt Lite access, which is being released in waves.
- The research-preview rollout or model availability has changed since launch.
Before troubleshooting further, check Bolt’s live account and agent settings because Forge is explicitly a research preview and access conditions can change quickly.
12
Should you duplicate a project before using Forge?
For serious projects, yes. That recommendation comes directly from Bolt.
Because Forge’s open models are experimental inside Bolt, duplicating the project gives you a simple rollback path if the agent makes broad changes you do not want to keep.
13
Does Forge replace Standard and Max?
No. Forge sits beside them.
You can switch between agents, and Bolt says Standard and Max retain their own allocations and privacy behavior. Forge is best understood as an additional experimental lane rather than a replacement for Bolt’s premium agents.
14
What happens after October 14, 2026?
Bolt says Forge will continue as an experimental open-model lab after the initial research-preview window.
What is not yet defined is the permanent allocation after the temporary up-to-50X Pro promotion ends. Recheck the live Bolt pricing and Forge documentation before relying on the launch-period allowance for future budgeting.
15
Bottom line
Using Bolt Forge is straightforward once you understand the access and consent model: choose Forge from the agent picker, review the opt-in, start with the default GLM model, build, and watch the separate Forge usage bar.
The most important practical rule is to treat Forge as experimental. Duplicate important projects, keep production-critical work in Standard or Max when appropriate, and remember that the large preview allocation is temporary rather than a guaranteed permanent plan feature.
Sources
Primary and supporting sources
Use the linked documentation to check current pricing, access and product behaviour. The dates above describe this article; they do not guarantee that every vendor claim remains current.