Overview
Connect Claude to read-only runtime evidence from your LogForge deployments and explicitly shared .logforge reports. Investigate errors, slow requests, container health and telemetry gaps with scoped logs, metrics, inventory, alerts, operational graphs and approved file inspection. Use feature tags to follow JavaScript, TypeScript, browser and Python diagnostics across frontend, backend and database boundaries. Shared reports show named checks, source excerpts and reached, filtered or unreached expectations, helping you choose a fix and verify the next run. Sign in to a LogForge gateway account, pair an outbound relay with your authorized LogForge instance, and explicitly share any report projects. Central credentials remain on the relay machine; account ownership, read-only permissions and LogForge entitlements remain enforced. The companion LogForge plugin adds setup and coding guidance for local diagnostics, Docker telemetry and selected remote Linux agents in Claude Code or an authorized executor. Claude web reads connected evidence; it cannot execute your local project, install Docker or deploy agents. From plugin 0.2.0, minimal local publisher aggregates default to ON for fresh Claude reporter state; saved disables remain off. Run python <plugin-root>/scripts/logforge_usage.py disable for a persistent off switch, including before first use. Source, prompts, log contents, feature tags and credentials are excluded. Hosted query processing is described separately in the privacy notice. These aggregates do not measure marketplace installs.
