DevSpeak's FREE plan lets you connect your own AI provider key, which removes the hourly request limit that applies to free accounts using DevSpeak's own models. Supported providers include OpenAI, Anthropic, Gemini, OpenRouter, and other OpenAI-compatible APIs that DevSpeak can reach, such as DeepSeek, Mistral, and Cerebras. Creating an account requires no credit card, and the optional 30-day VIBECODER trial requires no payment method and returns to the FREE plan at the end without a charge.
Do More on the DevSpeak FREE Plan: Connect Your Own Key
Most free tiers are a countdown. You get a fixed number of runs, you hit the wall mid-task, and the tool stops being useful exactly when you were starting to trust it. DevSpeak's FREE plan has a limit too — but it also includes the one setting that lifts it.
Two Ways to Translate on the FREE Plan
Every DevSpeak account can turn an informal requirement into a formal, audience-specific technical specification. On the FREE plan you choose how that translation is powered.
Use DevSpeak's models. Sign in and start translating. Nothing to configure, nothing to pay for, no key to manage. FREE accounts get a set number of translations per hour — enough to evaluate the product properly, not enough to sit inside a refactor all afternoon.
Connect your own key. If you already hold an API key from an AI provider, add it in settings and DevSpeak routes your translations through it instead. The hourly limit no longer applies, and DevSpeak does not meter that usage against any monthly allowance. You are working directly with your provider, on the quota you already pay for.
The second option is the interesting one, because it is usually assumed to be a paid feature. It is not. Bring-your-own-key is available on FREE, and it is why the FREE plan holds up under real use rather than functioning as a demo.
Setting It Up
Everything lives at Settings → Keys & Providers.
Presets cover the providers most developers already have: OpenAI, Anthropic, Gemini, OpenRouter, and an Other OpenAI-compatible option for providers such as DeepSeek, Mistral, Cerebras, NVIDIA, and Alibaba Cloud Model Studio. DevSpeak calls your provider from its own servers, so it accepts only endpoints on its allowlist; a model server on your own machine works only if you self-host DevSpeak. Choosing a preset fills in the endpoint, which you can edit. The model is yours to enter: it is a required field, because only you know which models your key can reach (Gemini is the exception: DevSpeak sets the model for you).
Three things are worth knowing before you save:
The key never comes back. It is encrypted on DevSpeak's servers, and the settings page only ever shows a masked placeholder afterward. There is no reveal button. Keep your own copy.
Attachment support varies by provider. Document and image attachments work on the Anthropic and Gemini routes. If you pick a provider that cannot accept images, DevSpeak warns you in the form before you save rather than failing on your first upload.
Compatible endpoints are trusted as configured. DevSpeak cannot detect what every OpenAI-compatible model supports, so if yours does not accept images you will see your provider's own error when you try.
Choosing Per Translation, Not Per Account
Connecting a key is not a mode switch. Once a credential is active, your provider appears as an extra option in the model selector inside the translation editor, alongside DevSpeak's own models.
Selecting your own provider sends that translation through your key. Selecting one of DevSpeak's models sends it through DevSpeak, subject to the FREE plan's hourly limit. You can switch between them freely, which is more useful than it sounds: running the same requirement through two different models is a direct way to work out which one suits your writing.
One detail catches people out. Choosing a DevSpeak model overrides your connected key for that translation. That is deliberate — picking a model from a list should do exactly what it says rather than quietly routing somewhere else. When you want your own provider, select it explicitly.
Where the Time Actually Goes
Connecting a key is not free in the literal sense; you are paying your provider. The case for doing it is that the setup cost is paid once, while the friction it removes is paid per task.
The workflows where that compounds are the ones with a translation in the inner loop rather than at the edges. Working through a backlog where every ticket needs restating for a different reader. Rewriting an evaluation section for the engineer who will be paged about it at 3 a.m. Regenerating a specification after the third requirement change of the day — where the whole value of the tool is that regenerating is cheap. An hourly cap is a real ceiling for all of those. Your own provider quota generally is not.
No Card, and Nothing Expires Into a Charge
Creating an account is a plain signup. The button says Get Started Free, and it puts you on the FREE plan.
The 30-day VIBECODER trial is separate and opt-in. You click Start free trial, and that is the entire flow — no payment method is requested at any point, because none is required.
What happens on day 31 matters more than what happens on day one: the trial ends and your account returns to the FREE plan. It does not convert into a subscription, and there is no charge, because DevSpeak never took a payment method to charge. Your connected provider key stays attached and keeps working exactly as before.
Trying VIBECODER for a month therefore costs nothing and risks nothing, and the plan you land back on is one that still does useful work.
What Upgrading Is Actually For
Since bring-your-own-key is already on FREE, the reason to upgrade is not throughput. It is the features where DevSpeak does work beyond passing your text to a model.
VIBECODER ($9.99/mo) adds document translation, iterative refinement, custom instructions, memory context, file attachments, and Context Projects.
DEVELOPER ($19.99/mo) makes API, CLI and MCP use unlimited (FREE and VIBECODER include one API key with 50 requests a month on the default model, enough to try the CLI or connect the MCP server), and raises the limits across the board — larger documents, more refinements, more Context Projects. This is the plan that makes DevSpeak scriptable inside CI.
Refinement is the clearest illustration of the split. It is the loop where you send a generated specification back with feedback and get a corrected document, and it is unavailable on FREE no matter whose key is connected — because refinement is DevSpeak's own orchestration, not a single model call. Connecting a key lifts DevSpeak's rate limit. It does not unlock DevSpeak's features.
Start Here
Create an account. If you already hold a provider key, spend ten minutes at Settings → Keys & Providers. You will know within an afternoon whether audience-specific translation belongs in your workflow — and you will find out without a countdown running in the corner.