Eversynced launches Pheebs, open-source AI coding telemetry

Pheebs walks a careful line on privacy. It records interaction patterns — the shape of a session — but explicitly never stores source code, file contents, file paths, directory structures, prompt text, command strings, your name, or your email. An engineer is identified only as an id behind a Pheebs token, or a truncated hash of their git email.

The tool can run two ways. In a self-hosted setup, you run the backend and telemetry goes from developers' machines to your own infrastructure; Eversynced never sees it. In the managed setup, the same open-source client points at a backend Eversynced operates, with reporting and dashboards layered on top as part of their AI Enablement Assessment — a 30-day telemetry sprint that ends in an executive debrief and a plan for the gaps.

On top of the raw telemetry, Eversynced proposes an AI Proficiency Model with two layers. Layer 1 reads repertoire across 24 practices in six competencies: models, artifacts, MCP, evals, context management, and orchestration. Layer 2 reads judgement signals such as verification coverage and model-fit rate. Eversynced says it instruments every one of its own engineers with Pheebs.