# EvidenceLab > Public beta decision-support workspace. The synthetic example can be explored without an account or API key. This document is public product information, not a live investigation or a source of empirical business results. ## Public resources - [Interactive workspace](/): Requires JavaScript. Initial HTML includes a product overview for clients that do not execute scripts. - [Public product guide](/about): Full HTML explanation, no JavaScript or authentication required. - [Service health](/api/health): JSON version and provider metadata. - [Provider choices](/api/providers): JSON provider/model choices. - [Synthetic demonstration inputs](/api/demo): Fictional SaaS scenario; all numeric estimates are illustrative assumptions. ## Current capabilities Define objectives, baselines, targets and horizons; map variables and relationships; record evidence and structured constraints; compare candidate actions; inspect feasible, infeasible and unknown results; provide conditional recommendations or evidence gaps; export JSON or Markdown and import saved JSON. Controllability, permission and fixed status are independent. Variable memberships may overlap. Source labels are supplied by users and are not independently verified. Hypothetical assumptions require explicit approval for conditional comparisons. ## Optional AI providers OpenAI, Gemini and Groq are supported as visitor-supplied API-key choices. A key is not required for manual editing or the synthetic example. Provider authentication, quotas and charges depend on the visitor's account. Live generation across all providers remains to be independently verified. The application server receives a key transiently during a provider request. Keys are excluded from browser saving and report exports. No operator-funded default key is supplied. Provider and infrastructure data policies still apply. ## Access and limitations Anonymous public access is enabled. Some external tools may restrict this hosting domain or lack JavaScript execution. The product guide and this document provide readable alternatives, but cannot override another service's network policy. The product does not guarantee outcomes, establish causal effects, independently verify evidence or claim a universal optimum. Sprint scheduling, adaptive cognitive-load measurement and automated goal discovery are proposed future work. There is no persistent server investigation database. This llms.txt file supplies optional discovery information. It does not guarantee indexing, access by any particular LLM, or automatic interaction with the application. The multi-question survey is disabled. After their own guided comparison, visitors can explicitly submit one optional usefulness answer. It excludes investigation content and contact details. No automatic usage analytics is collected. Receipt-based feedback deletion remains available; see /about#feedback. Experiences: / is the guided start, with a synthetic example and a short manual outline. /workspace is the detailed decision workspace. The older /v0 and /v1 addresses remain available. Optional AI is a deliberate choice after the first result. Provider keys and charges remain each visitor's responsibility. Guided outlines do not fabricate outcomes, grant permissions or claim an optimum. No assessment questionnaire or automatic usage analytics has been added.