Foretop - CI-native drift scanners for AI systems

Hi Product Hunt — I’m Nicolas, the maker of Foretop. I built Foretop after noticing that some of the most consequential changes in an AI system happen outside its repository. A provider retires a model, an API route ships without declared telemetry, an MCP server gains a capability, a dependency brings a licence obligation, or a vendor changes its subprocessors. Ordinary linters cannot see those changes. Foretop turns them into one evidence-backed CI workflow: - Ebb — model retirements: `uvx foretop-ebb scan .` - Telltale — OpenAPI telemetry coverage: `uvx foretop-telltale check .` - Charter — MCP capability drift: `uvx foretop-charter scan .` - Lading — AI asset licence obligations: `uvx foretop-lading scan .` Four Apache-2.0 scanners run locally or in CI with no account, and source stays local. Undertow is scheduled monitoring for vendor subprocessor changes rather than an installable CLI. Every result carries evidence, and a check that could not establish an answer stays unknown instead of becoming clear. The public reports are independent analyses of real open-source repositories at pinned commits, not endorsements, and the same commands reproduce them. Foretop Hub is the optional shared layer for teams that want run history, repository policies, ownership, expiring suppressions, comparisons, and evidence exports. Sending finding metadata is explicit; repository source bodies are not uploaded. Free supports 3 repositories with 30 days of finding detail. Team supports 10 repositories with 365 days of finding detail for €39/month or €390/year, including tax. Run history remains available on both plans. I’d especially value feedback on which kind of external drift is hardest to catch in your stack, whether Foretop’s findings make the next action clear, and what would stop you adding one of the checks to CI.