TL;DR. Bring your own AI (BYO AI) means you connect the AI you already use and pay for (Claude, ChatGPT, Cursor, Codex) to a platform, instead of buying that platform's AI credits. The platform handles hosting, database and deploy; your AI handles the building. Sometimes called bring-your-own-subscription AI, it took off once MCP (Model Context Protocol) made cross-tool AI connections standard in 2024-2025. The practical benefit: the platform's free tier doesn't have to meter AI, because it isn't paying for AI. (Updated August 2026.)
The definition
BYO-AI describes any platform or tool where:
- You use an AI tool you already have (or separately choose): Claude Code, Cursor, ChatGPT, Codex, etc.
- The AI tool connects to the platform via an open protocol (usually MCP).
- The platform exposes its capabilities (deploy, database, auth) as tools the AI can call.
- The platform doesn't bundle, meter, or charge for AI inference, because it doesn't provide AI.
The term itself is an analogy to BYOD (bring-your-own-device) in corporate IT, or BYOL (bring-your-own-license) in enterprise software. The idea: the bundle is unbundled, and each piece is sourced separately.
Why BYO-AI emerged
Three forces came together in 2024-2025:
1. AI inference is expensive at platform scale
If a platform gives every free user AI as part of the bundle, the AI is the dominant cost. Platforms that do this (Lovable, Bolt.new, Replit's agent, v0) have to meter something (credits, daily tokens, monthly messages) to keep the math working. The free tier isn't truly free; it's a limited allowance.
A BYO-AI platform doesn't pay for AI inference. Hosting a small web app with a Postgres database costs the platform a few cents per month per user. So the free tier can stay genuinely free with no metered ceiling on iteration.
2. Users increasingly have their own AI subscriptions
By early 2025, millions of users had Claude Pro, ChatGPT Plus, Cursor Pro, or similar. For those users, paying a second time for bundled AI inside a builder platform was paying twice. BYO-AI lets them use what they already have.
Even users without a paid subscription often have access to free tiers of the major AI tools. Claude free, ChatGPT free, and Cursor free are all sufficient for building small-to-medium apps on a BYO-AI platform.
3. MCP made it technically feasible
Before MCP, "bring your own AI" would have meant custom integrations for each AI tool: one for Claude, one for ChatGPT, one for Cursor, etc. The engineering cost was prohibitive.
MCP (Model Context Protocol), introduced by Anthropic in late 2024 and now supported by most major AI tools, made the "bring any AI" promise real. A platform builds one MCP server and every MCP-compatible AI client can use it.
How it works in practice
Walk through a concrete BYO-AI workflow on a hypothetical hosting platform:
- Install an AI tool. For example, Claude Code:
npm install -g @anthropic-ai/claude-code. - Connect the hosting platform. Platforms typically either sign you in with OAuth from inside the AI tool, or issue a token you paste into its config.
- Register the platform's MCP server with your AI tool. One line of config: a command (
claude mcp add ...), a JSON entry (Cursor), or a TOML block (Codex). - Talk to the AI. "Build a habit tracker and deploy it." The AI calls the platform's tools (
create_project,write_file,deploy) to build and ship. - Get a live URL. The platform returns the deploy URL. Your app is running.
The platform only sees tool calls; it doesn't pay the AI provider for the conversation. You (or whoever's paying for your AI subscription) does.
BYO-AI vs. bundled-AI platforms
| BYO-AI | Bundled AI | |
|---|---|---|
| Pricing model | Platform charges for hosting features; AI has its own price | One bundled subscription for AI + hosting |
| Free tier | Often genuinely free forever (no AI to meter) | Credit-based; usually has monthly limits |
| Setup | Requires connecting an AI tool | Ready to use out of the box |
| AI flexibility | Switch AI tools freely; mix multiple | Stuck with the platform's choice |
| If AI improves elsewhere | You get the benefit (just switch your AI tool) | You wait for the platform to adopt it |
| Who pays for AI | You (directly to the AI provider) | You (via platform, which passes costs through) |
BYO AI vs credits: the three billing shapes
"Bundled" hides two different things, so it is worth separating the three ways an app-building platform can bill for the AI part. General characterisations based on public information as of August 2026; check each vendor for current details.
| Shape | Who meters the AI | What you pay | Where you see it |
|---|---|---|---|
| Credits | The platform | A plan that includes a monthly allowance of credits, messages or tokens, with top-ups when it runs out | Most bundled AI app builders as of August 2026; Lovable, Bolt, Replit's agent and v0 all publish credit or message allowances, and are genuinely the fastest way to a first prototype |
| Bring your own key (BYOK) | The model provider, per token | A platform fee plus an API bill from OpenAI, Anthropic or Google that rises and falls with usage | Many developer tools, and some builders offer it as an option next to credits |
| Bring your own AI (subscription) | Nobody new: your existing AI subscription's own limits apply | A flat platform fee (or $0) plus the Claude, ChatGPT, Cursor or Codex plan you already have | Hatchable; DIY stacks where your coding agent deploys to hosting you rent |
Credits are the simplest to start with and the easiest to misjudge: the meter runs while you iterate, and the price of a unit of work is opaque until the pack is gone. BYOK is the most transparent per token and the most variable per month, and it means keeping an API key safe. Bring your own AI is flat and predictable on the platform side, and it leans on something most builders already own, an AI subscription; its costs are the one-time setup step of connecting the tool, and the fact that your AI tool's own usage limits still apply.
Hatchable is the third shape on purpose. There are no credit packs, no message credits and no per-prompt fees; the person's own AI subscription writes the code over MCP, and the platform charges a flat rate for hosting: the free plan is $0 with unlimited private projects and one published app, no card, and Builder is $12 a month for unlimited published apps. If you want to see the connection step, the Claude Code, Cursor and Codex guides each take about two minutes.
When BYO-AI makes sense
- You already have a paid AI subscription (Claude Pro, ChatGPT Plus, Cursor Pro). Paying again through a bundled platform is paying twice.
- You want to use a specific AI tool, maybe because you love Cursor's interface, or Claude's strengths in long-form reasoning. BYO-AI lets you pick.
- You want to iterate heavily without worrying about credits. Bundled platforms meter; BYO-AI platforms don't.
- You want a long-lived project. BYO-AI's economics (cheap hosting plus AI you already pay for) are more durable at the year-plus timescale.
- You want to be able to switch AI tools if a better one comes along. Without BYO-AI, you're tied to the platform's choice.
When bundled AI makes sense
- You don't have an AI tool and don't want to set one up. Bundled platforms are faster to start.
- You'd use AI only for app building. If AI has no other use in your life, the bundled subscription isn't redundant.
- You want one subscription to think about. Simplicity has value.
- Your project is ephemeral. If you're making a prototype that'll be thrown away, paying the bundled subscription for a month is fine.
The BYO-AI platforms
As of August 2026, BYO-AI platforms for app hosting are a small category. The main players:
- Hatchable: MCP-native app hosting; works with any MCP-compatible AI tool.
- Self-hosted stacks: Cursor + Supabase + Cloudflare Pages, or similar DIY combinations. Technically BYO-AI, but you're the one stitching everything together.
Expect this category to grow. Once MCP became the standard, the economic logic of BYO-AI started to make obvious sense for hosting platforms specifically: hosting is a commodity that can be free forever; AI isn't. Unbundling makes the hosting half scale.
BYO-AI vs. self-hosting
A fair question: if you're bringing your own AI, why not self-host everything? You could run your own Postgres on a VPS and have Claude Code deploy to it.
The answer is the same as "why use any hosting platform": setup, maintenance, backups, scaling, TLS certificates, DNS. Self-hosting is free dollars but costs time. BYO-AI hosting platforms are free dollars AND free time, because someone else handles the operations while you focus on the apps.
For developers who enjoy infrastructure work, self-hosting can be more rewarding. For people who just want apps to work, a BYO-AI platform is the middle ground: you keep the control and portability of self-hosting (standard Node + Postgres, exportable) while not having to run the servers.
The future
A prediction, labeled as such: BYO-AI will become the default for long-lived app hosting over the next 2-3 years. Bundled platforms will remain popular for prototypes and zero-setup experiences, but "I want my app to still be here in three years" users will increasingly use BYO-AI platforms.
The reasoning: hosting costs are a commodity that free tiers can absorb; AI inference costs aren't. A platform that structurally can't give away the expensive part (AI) will always have to meter it. A platform that can (BYO-AI) won't. The users who care about "still free in three years" will migrate toward the structure that supports it.
Try BYO-AI hosting.
Hatchable is free forever for apps your AI builds. Bring Claude, Cursor, or any MCP-compatible tool.
Get started free →Frequently asked questions
What does BYO-AI stand for?
Bring-your-own-AI. A model where a platform doesn't provide or charge for AI: users connect their own AI tool (Claude, Cursor, ChatGPT, etc.) to the platform via an open protocol like MCP.
Do I need a paid AI subscription to use a BYO-AI platform?
No. Free tiers of Claude, ChatGPT, and Cursor are all sufficient for building on BYO-AI platforms for light to moderate use. A paid AI subscription helps if you iterate heavily or build multiple apps, but it isn't required.
Is bring your own AI cheaper?
It depends on how much you build. If you already pay for an AI subscription, or get by on a free tier, BYO AI is usually cheaper, because you stop buying a second AI allowance and the platform fee is low or nothing (Hatchable's free plan is $0 with one published app; Builder is $12 a month). If you have no AI subscription and build rarely, a bundled plan's single bill can be simpler and roughly comparable. Heavy iteration tends to favour BYO AI, since credit packs run out faster than subscription limits, but plans and limits change often, so compare your actual month: what the AI costs you today, plus what the platform charges.
Can I use multiple AI tools with a BYO-AI platform?
Yes. MCP-compatible platforms let you connect any MCP-compatible AI tool. You can use Claude for one session, Cursor for another, and both write to the same project.
What happens when my AI tool's free tier runs out?
The BYO-AI platform is unaffected: it doesn't know or care about your AI tool's limits. You can either upgrade your AI tool to its paid tier or switch to a different AI tool (some have more generous free tiers than others). The platform keeps working.
Is MCP the only way BYO-AI platforms work?
Currently, yes: MCP is the protocol that made cross-AI-tool connectivity viable. In principle, other protocols could emerge, but MCP has enough momentum (supported by Anthropic, OpenAI, Cursor, Google, and independent tools) that it's the de facto standard.