Beta Access

We're a tiny, self-bootstrapping team building knowledge infrastructure that tries to matter. No venture capital, no pressure to scale before we're ready, no shortcuts on quality. Just a small group who believes deeply in user-owned data and wants to build this right.

That means during beta, we're have to be a bit more careful than your average startup. We can't afford to scale recklessly – both financially and architecturally. Instead, we're inviting people in batches, learning from each cohort, and making sure the infrastructure can handle growth before opening the floodgates.

---

How Beta Access Works

Recurse is currently in closed beta with a waitlist. We're inviting projects and individuals in batches as we validate capacity and refine the system. The waiting period isn't artificial scarcity—it's genuine constraint. We're building infrastructure that needs to work reliably before hundreds or thousands of people depend on it.

When you apply through the homepage, you'll choose between two access levels: full developer access (API + UI) or UI-only access. This distinction matters because API usage puts more load on our infrastructure than dashboard interactions. If you primarily want to use the chat interface and dashboard, we can expedite your request. If you need programmatic access for building integrations or automations, there's typically more of a wait.

We process applications in batches. Sometimes weekly, sometimes bi-weekly, depending on current capacity and what we're learning from existing users. Priority goes to people with specific projects, interesting use cases, or who are willing to provide thoughtful feedback. We're especially interested in researchers, writers, developers building AI applications, and anyone exploring ideas that could push the system in unexpected directions.

---

Bring Your Own Key

Here's the other honest part: we can't pay for millions of AI processing requests. When you process documents in Recurse, we use your AI provider key to extract semantic structure and create the knowledge graph. The embeddings—which power search and basic retrieval—we handle ourselves. This means core RAG functionality always works, even if you run out of AI provider credits.

This keeps Recurse free during beta while we figure out sustainable pricing. You only pay for the semantic extraction (frames, relationships, summaries), and your knowledge graph lives in our infrastructure with full search, navigation, and retrieval capabilities at no additional cost. Most users spend less than $10 per month on AI processing, but it scales with your usage—process more documents, pay more to your AI provider.

Recurse works with any OpenAI-compatible provider. Get an API key from OpenAI, Anthropic, DeepSeek, or OpenRouter, add it to your settings, and start processing. Your keys are encrypted at rest and only used to make AI requests on your behalf.

---

Access Levels Explained

UI-Only Access means you can use the dashboard, upload documents through the interface, chat with your knowledge, and explore the graph view. You won't have API keys or programmatic access, but you get full functionality for personal knowledge management and exploration. We can typically expedite these requests because dashboard usage doesn't put as much load on our infrastructure.

Full Developer Access includes everything in UI access plus API credentials for programmatic integration. This lets you build applications, automate uploads, integrate with existing tools, and use Recurse as infrastructure for your own projects. API access requires more careful capacity planning on our end, so there's usually a longer wait—but it unlocks the full power of what Recurse can do.

You can always upgrade from UI-only to full developer access later. Just let us know what you're building and we'll work with you.

---

Getting Started After Approval

Once you're approved, you'll receive an email with your dashboard access (and API credentials if applicable). From there:

---

Why We're Doing It This Way

We could raise venture capital, scale aggressively, and figure out sustainability later. But that path leads to compromises we don't want to make. Selling user data to cover costs. Optimizing for engagement metrics instead of utility. Building features investors want instead of what users need.

Instead, we're building slowly with user ownership as the foundation. Your knowledge stays yours. Your data isn't training material for our models. Your usage doesn't subsidize free tiers for others. This approach requires careful scaling—hence the waitlist—but it keeps incentives aligned.

Bring-your-own-key means users who process more pay more, which feels fair. It also means we're not incentivized to push unnecessary AI usage just to justify our infrastructure costs. The system only processes what you explicitly ask it to process.

The waitlist ensures we can maintain quality as we grow. Each batch of new users teaches us something: how people actually use the system, what breaks under load, where documentation needs improvement, which features matter most. We iterate between batches instead of trying to support exponential growth before the foundation is solid.

---

The Future

We're exploring sustainable funding models that don't compromise on ownership or privacy. Options include paid team plans for organizations, compute credits for power users, research/education partnerships, and possibly grant funding for public infrastructure. If you're interested in sponsoring compute costs or have ideas about sustainable funding, reach out at hello@recurse.cc.

Until then, the waitlist and bring-your-own-key approach lets us keep building without venture capital pressure. We answer to users, not investors. The code will eventually be open source. The knowledge you build is yours to export anytime.

We're grateful you're here during the early days. Let's build something that lasts.

---

Common Questions