ProofHound Docs are now live

ProofHound now has a public documentation home for teams evaluating the product, running their first optimization, integrating it into their own systems, or deciding whether to self-host. The goal is simple: keep the important product knowledge outside chat threads and release notes, and make every common path easier to follow.

What the docs cover

• Product orientation: why ProofHound exists, how the core objects fit together, and how hosted organizations, projects, and access work.
• A first-run tutorial that walks from a labeled dataset to a measurable prompt improvement through the automated optimization loop.
• How-to guides for connecting a model, importing a dataset, creating a prompt, targeting a specific metric, publishing a prompt, using access tokens and MCP tools, setting up connectors, and annotating release results.
• Explanation pages for the optimization loop, prompt versions, connectors, releases and canaries, and why overall accuracy is often not enough for high-stakes classification.
• Reference pages for dataset formats, model configuration, prompt anatomy, optimization settings and metrics, release behavior, roles and permissions, API access, and MCP tools.
• A self-hosting guide for teams that want to run the open-source edition on their own infrastructure with Docker Compose.

What comes next

The docs are a living surface. As ProofHound Cloud and the open-source project keep changing, we will keep tightening the current pages, adding screenshots and examples where they help, expanding troubleshooting notes, and keeping Chinese and English documentation aligned.

We will also use the docs to make new capabilities easier to adopt. As connectors, API flows, MCP tools, billing limits, and future optimization modes grow more capable, their contracts and recommended workflows will land in the documentation first, so teams can move from evaluation to production with fewer surprises.