> ## Documentation Index > Fetch the complete documentation index at: https://docs.getmillwork.dev/llms.txt > Use this file to discover all available pages before exploring further. # Where Solver fits > See where governed, verifiable AI execution changes the way a team builds. Solver is most useful when AI work has outgrown one hard-coded call and still needs to remain explainable. ## Build across providers with one run contract **The situation:** an application needs access to models served by different providers, clouds, or gateways. **What changes:** provider access is configured separately. The live catalog shows tested model deployments that may be registered. The application submits one run contract over its registered options. **What you can inspect:** on deployment-bound model runs, requested, resolved, and provider-reported model identity, serving provider, access path, recorded usage, and the selected route. ## Add safeguards to agent work **The situation:** an agent workflow can call models, services, or procedures, but each path applies limits and failure handling differently. **What changes:** registered capabilities share data, cost, time, readiness, and verification boundaries. Solver takes only a fallback or repair action the run permits. **What you can inspect:** eligibility, route rationale, output check, bounded action, and final outcome. ## Change supply while keeping product code stable **The situation:** a provider route changes, a new account becomes available, or one route is no longer ready. **What changes:** the capability and its tested route live behind Solver rather than inside the feature's request code. **What you can inspect:** for deployment-bound model runs, the provider and model that answered, including a more specific model identity reported by a gateway when available. ## Keep operational proof and customer content separate **The situation:** a platform team needs to understand what ran while keeping prompts and outputs separate from the audit artifact. **What changes:** the content-bearing result and content-free receipt are separate records joined by the run ID. **What you can inspect:** lifecycle, route, recorded usage, checks, bounded actions, and, on deployment-bound model runs, provider identity. Task and output content remain separate from the receipt. ## Private-preview fit The current preview is designed for teams evaluating governed model runs over REST with public information or test data created for the evaluation. Check the availability page before planning restricted-data or production commitments. Turn your requirements into a short, evidence-based evaluation. See what is available now and how the release can expand.