Pick the tier that fits
your research.

Yes. A 14-day free trial will be available at launch. The trial covers Essential-tier features. That's enough to run the full strategy-building, backtesting, and analysis workflow end-to-end so you can evaluate the platform properly before committing. You can upgrade to a higher tier at any time during or after the trial.

You can cancel from your AlphaFind account on the website. That's where your subscription, billing history, invoices, and payment method are managed. You can also cancel from within the app itself. Either way, your cancellation takes effect at the end of your current billing cycle, and you keep paid-tier access right up until then.

Your research stays with you. AlphaFind is a native desktop application. Your strategies, datasets, and research records live on your own machine, not on our servers.

If you cancel your subscription but keep AlphaFind installed, everything stays exactly where it is. You just can't log in until you resubscribe, at which point you pick up where you left off.

If you want to uninstall, export is built directly into AlphaFind. You can export your strategies, datasets, and research records straight from the app itself, on your own schedule. No support ticket, no waiting on us, no hoops. Your research is yours.

Yes. If you want to move up to a higher tier mid-cycle and unlock its features immediately, we prorate fairly. The unused portion of what you've already paid is credited toward the new tier, and you only cover the difference through to the end of your current cycle. Downgrades take effect at the start of your next billing cycle.

Your perpetual license includes 12 months of feature updates from the date of purchase. After that, the software stays fully functional at whatever version you have. It doesn't stop working or expire. You just stop receiving new features until you choose to upgrade.

When you're ready for newer features, or want to move up to a higher perpetual tier (Pro → Ultimate), you can do so at a 30% discount within two years of your original purchase. The upgrade refreshes your update window and brings you current with the latest release.

AlphaFind is a native desktop application. At launch we'll support:

macOS: Sonoma (14) or later, Apple Silicon (M1 or newer). Intel Macs are not supported.
Windows: Windows 11, 64-bit, on x86-64 (Intel Core i5 10th-gen or newer / AMD Ryzen 5 3000-series or newer). Workstation-class Intel Xeon and AMD EPYC CPUs are fully supported.
Linux: Ubuntu 24.04 LTS or later, x86-64 with the same CPU minimums as Windows.

Recommended hardware: 32GB system memory, a dedicated GPU with Metal, DirectX 12, or Vulkan support (Apple Silicon integrated GPU counts), and a modern mid-range CPU. You can run on less, but larger datasets, long-running optimisations, and scientific workflows benefit from more RAM and a faster GPU.

We don't offer general refunds. That's what the 14-day free trial is for. Use it to evaluate AlphaFind end-to-end before you commit.

If you're running on hardware that meets our minimum specs and AlphaFind isn't working as it should, our support team will work with you until it does. That's our commitment to paid subscribers.

Refunds for reasons outside that (change of mind, didn't end up using it, paid and forgot) aren't available. The trial window is there so you know exactly what you're buying.

Yes. Point the LLM at AlphaFind's developer documentation first. That's the manual for how strategies must be written. Once it's read the docs, describe what you want ("a mean-reversion on S&P 500 constituents using a 20-day Z-score entry and a 5-day exit"), and it produces Python you can paste straight into AlphaFind.

Strategies are written in Python, but structured in a specific shape so our engine can compile them to optimised low-level code for fast backtesting. The documentation tells the LLM exactly what that shape looks like (syntax rules, available primitives, data access patterns), so the output pastes in and runs.

Practical workflow: give the LLM the docs, describe the idea, paste the result. No Python skills required on your end. The LLM handles the conformance.

Basic: four fundamental methods. Grid Search and Random Sampling for parameter tuning. Walk-Forward for out-of-sample testing. Monte Carlo for robustness. Enough to run a credible research workflow on a single strategy.

Standard: everything in Basic, plus four more. Bayesian optimisation and Simulated Annealing for smarter search over large parameter spaces. K-Fold Cross-Validation and CPCV (combinatorial purged cross-validation) for ML-grade rigour.

Advanced: everything above, plus the full evolutionary suite. Genetic Algorithm, Differential Evolution, Particle Swarm, and CMA-ES. Population-based methods for complex, high-dimensional search.

Full methodology documentation ships with the product.

Scientific workflows turn loose backtesting into a structured research process. Every investigation runs on a chain: hypothesis → protocol → expected outcome → experiment → result → interpretation → next step. You declare what you believe and how you'll test it before running, and what actually happened gets written to the record afterward.

Pick the rigor level that fits the work. Sandbox is free exploration for early ideas. Soft Science keeps the structure (journals, linked evidence, expected-vs-observed tracking) but lets you revise and iterate. Hard Science is the strongest mode: holdout data is fully protected, you validate against it once, and that result becomes a permanent, incorruptible part of the research record. The constraint is the point. It stops you from probing the holdout until you get the answer you wanted.

Everything flows into a research journal and findings catalogue. Every hypothesis, caveat, dead end, and conclusion is searchable and linked to the strategies, cohorts, graphs, and pivots it came from. Over time your research becomes a knowledge base you can return to, not a scratchpad.

Yes. Every tier includes custom data import. Bring your own price history, fundamentals, or alternative datasets. The built-in import tool handles schema detection, timestamp parsing, OHLCV recognition, validation, and repeat-import templates so onboarding your own data doesn't feel like a chore. Larger bulk pipelines and automated feed ingestion are on the roadmap for later stages.

AlphaFind is designed to work equally well for serious retail quants and working professionals. Same product, same feature depth. The tier structure reflects how much research you're running, not who you are.

Organisation-level controls are on the post-launch roadmap: multi-seat accounts, SSO/SAML, audit logging, role-based access, and governance over shared compute and data entitlements.

If you're evaluating AlphaFind for a team, a desk, or a specific institutional use case, we'd genuinely like to hear about it. Write to hello@alphafind.ai and you'll get a reply from a human.

At launch, AlphaFind is bring-your-own-data. Import your own price history, fundamentals, or research datasets through the built-in import tool. Every tier includes it. This keeps the platform affordable at launch and lets you use the data you already have or already license.

Managed historical data arrives in a later stage, on an à la carte basis. You'll be able to add specific asset classes and depth as paid options rather than pay for a blanket feed bundle. The full menu is published as vendor agreements firm up.

Live data feeds follow after that, added stage by stage as new data partnerships come online.

Broker integrations (Interactive Brokers and others) are on the roadmap but not in the initial launch. The first release focuses entirely on research-grade backtesting, validation, and analysis.

Execution connectivity follows once the research workflow is solid.

An integrated AI assistant is on the roadmap. Today the flow runs in the other direction: you generate strategy code in an external LLM and paste the Python straight into AlphaFind.

The in-app assistant arrives in a later stage, with awareness of your current workflow. Its role is to help you draft and debug strategy code, interpret backtests and metrics, reason about candidates and which validation path to take next, and recall relevant prior research from your own journals and findings. Later still, an optional local model tier for users with capable hardware.