Recommended Free Tools
drobek is an open-source hosting platform for small browser apps that you build with an external coding agent. Its core distinction is billing: drobek does not sell AI tokens, so model usage is paid to your agent’s provider, while hosting is a separate cost whether you use the hosted service or run your own server. The platform is described by its creator, Tomas Grasl, in a DEV Community post dated September 28, 2026.
The post opens with a question that many people who build small tools end up asking: “Where do the little apps your colleagues vibe code end up today?”
As an Amazon Associate I earn from qualifying purchases.
What drobek is built for
Grasl frames drobek around useful small apps that are easy to share, without requiring you to maintain a custom backend for every tool. The scope is deliberately narrow. drobek targets browser apps. It is not a general-purpose server runtime, and the author states that it is not a place to run arbitrary server code or background workers. If your project needs a long-running process or a custom API server, drobek is the wrong layer for it.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesHow the build-and-publish loop works
The workflow runs through a coding agent that you already use. Grasl names Claude Code, Cursor, and Codex as examples. The sequence he describes is:
- Connect the agent to drobek over MCP. The agent gets the tools it needs to create an app and submit files.
- Send the source files. The agent creates an app and submits its source for compilation.
- Address build errors. drobek compiles the files and returns any errors to the agent, which fixes them and resubmits.
- Inspect the preview. A successful build receives a preview URL where you can check the app before it goes live.
- Publish. After review you publish the build. Each successful build is numbered, so you can later publish an earlier version rather than only the newest one.
Who pays for what
The article’s central point is that AI spending and hosting spending are different line items. The agent’s provider bills for model usage. drobek does not resell tokens. Hosting still costs money, and the article makes no distinction on that point between the two deployment paths. In Grasl’s words:
“drobek doesn’t sell you AI tokens. Your model usage stays with your agent’s provider. Hosting still costs money, whether you use a hosted service or run your own server.”
The motivation was a wish to keep using a personal coding agent and deploy the result somewhere, without paying a platform for model tokens on top of hosting. The post gives no plan prices, usage limits, or cost comparison with other platforms, so any budget decision needs the current pricing from drobek directly.
What the article does and does not establish about cost
| Question | What Grasl’s post states | What is not stated |
|---|---|---|
| Who bills AI model usage | The agent’s model provider | Provider-specific rates |
| Who bills hosting | Hosting costs money, hosted or self-hosted | Any price, tier, or limit |
| Hosted option | A hosted version at drobek.app | Current plans, service terms, uptime |
| Self-hosted option | Users can run their own server | Hardware, deployment requirements |
| Source license | AGPL-3.0 | Not stated beyond the license name |
How the app runs
Grasl describes a split between browser code and platform services. App JavaScript runs in the visitor’s browser. drobek compiles and serves that code; it does not execute agent-written app code on its own server. Backend functionality comes from platform modules rather than from code the agent writes. According to the post, the modules cover:
- login
- data storage
- forms
- file uploads
- owner notifications
- API proxying
Each app has its own origin. Owners enter API secrets for the relevant modules in the drobek dashboard, and the module uses them server-side, so the coding agent never needs the secret values.
Security and data cautions
The author is clear that the platform does not remove the need for configuration and review. These points come from his own account and have not been checked by an independent security audit:
- Access rules are your responsibility. Data-access rules still need to be configured and reviewed. Hiding a button in the frontend does not enforce access control on the backend.
- Preview and production share data. Both use the same backend data for an app.
- Rollback does not reverse data changes. Reverting to an earlier build changes the code, not the database.
Those three cautions matter most if your app stores user records or handles payments-adjacent logic, because a rollback will not undo a bad write. Test schema-changing work against data you can afford to lose before publishing it.
Limits to weigh before you adopt it
- Browser apps only: no arbitrary server code and no background workers.
- Model usage is billed by your agent’s provider, so the cost of your coding sessions depends on that provider’s terms, not on drobek.
- Hosting cost exists in both the hosted and self-hosted paths, and the post gives no figures for either.
- The post describes the platform from its creator’s perspective; the availability, infrastructure provider, and deployment requirements for self-hosting are not stated in it.
Where this fits
drobek makes the most sense if you already work with a coding agent, your apps are browser-based, and you want the agent’s model bill kept separate from where the app is served. If you need server-side processes, or if you want a single bill that covers both AI and hosting, the design Grasl describes will not match your needs.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
