Hey,

Thomas here, co-founder of Raycast. Thanks for raising concerns about some upcoming pricing changes with v2, particularly around Custom Providers and BYOK. To answer the main question directly: yes, under the new v2 plan, BYOK will require Raycast Pro.

I can see why “pay us so I can use my own key” feels strange. Dianne’s response accurately described the plan, but we shared that decision without explaining the broader pricing direction or the product changes behind it. We should have communicated this better. I want to provide a bit more context and be clear about what is decided today and where we’re heading.

Why We’re Changing Pricing

We launched Raycast Pro in May 2023 as a paid subscription. As part of it, you got Raycast AI, Cloud Sync, custom themes, and extended Clipboard History. A big part of the Pro plan was access to all LLMs.

Since then, the AI landscape has changed significantly. There are many more models, their costs vary considerably, and agentic loops have increased usage substantially. We tried bundling the Raycast product and all inference into one flat subscription, but with growing token usage, that no longer seems sustainable. So we took a step back and looked at the value Raycast itself provides.

Raycast as the Interface Layer

Raycast has always been about building a better interface on top of your existing tools. We think the clearest principle is that Raycast should be the interface layer, not the inference layer.

We spend our time making that interface as useful as possible, from the Launcher through Dictation and Notes to AI Chat. With v2, we’re rebuilding our AI system, adding memory, projects, automations, and more capable ways for AI to complete work for you.

With that in mind, we want to separate the value Raycast provides from the underlying inference costs. The Raycast subscription pays for the product around the model: the interface, context, integrations, tools, and execution layer. How the intelligence underneath it is supplied should become a separate and transparent choice. I feel this aligns our incentives much better because our business succeeds by making Raycast better, not by maximizing a margin on tokens.

I also understand that changing an existing feature is frustrating and potentially disruptive, especially for people who have already built their workflow around BYOK, Custom Providers, or local models.

What This Means in Practice

At the launch of v2, there will be two confirmed ways to use AI through Raycast Pro:

• Built-in models: We’ll continue providing access to LLMs through Raycast for convenience. We’re also exploring a more transparent usage-based model, but the details aren’t final yet. I’m curious to hear which approaches you’ve seen work well.
• BYOK: You can bring your own key for OpenAI, Anthropic, or Google. We’ve seen many people use API keys provided by their employers.

Longer term, we want to offer more ways to supply the intelligence underneath Raycast:

• Existing subscriptions: Many people already pay for ChatGPT or Claude, and we don’t want to force them to purchase the same underlying AI capabilities again through Raycast. We’re working on ways to bring some of those existing subscriptions into Raycast.
• External and custom models: We’re working toward a way to connect more local and custom models. This is intended to become the path for Custom Providers in v2 and replace our existing Ollama and OpenRouter integrations. We currently see Custom Providers used mostly in enterprise environments, but I’d be interested to hear which other use cases matter to you.

All of these options use the same Raycast product around the model, which is why they will require a Pro subscription in v2. Your API key or existing subscription covers the LLM. Raycast Pro covers the interface, context, integrations, tools, and execution layer around it.

We offer a free 7-day Pro trial with no credit card required, so you can decide whether the product around the model is valuable to you. This is the current plan for v2, and we’ll publish the complete pricing before these changes roll out more broadly.

Going Forward

The industry is in a weird state. Many startups subsidize tokens to grow, but that isn’t the route we want to take. We see Raycast as a fundamental tool for many professionals and feel responsible for building a sustainable business that can control its own future and keep the product around for a long time.

For us, sustainability means charging for the value the product creates without betting on tokens becoming cheaper in the future. It also means being honest about the distinction between the model costs and the ongoing work required to build and maintain Raycast. This is why we chose a recurring subscription instead of a one-time payment. We ship updates every week and want the business model to align with that continued development.

This isn’t a plan to put all of Raycast behind a paywall. As before, the Launcher and the vast majority of the core Raycast experience will remain free.

Thanks again for holding us accountable. We should have explained the broader direction before communicating an isolated pricing decision. We haven’t figured out every detail yet, and I understand that this pricing model won’t work for everyone. But I hope the reasoning and where we’re heading are clearer now.

I’m here to listen and would love to hear your feedback, especially about how you use BYOK, Custom Providers, or local models today and what feels right to you.