endpoint-aiops v0.1.0 — preview

Endpoint AIops v0.1.0 — preview

Governed AI-ops for managed-endpoint fleets (thin clients / VDI) for AI
agents, with a built-in governance harness (audit, policy, token/runaway
budget, undo-token recording, graduated risk tiers) and an encrypted credential
store. Standalone — no external skill-family dependency.

Preview / mock-only. All behaviour is validated against mocked REST
responses; it has not been run against a live endpoint-management server. The
fastest live check is endpoint-aiops doctor.

Highlights

• 9 MCP tools (7 read, 2 write), every one wrapped with @governed_tool.
  • Read: fleet overview; inventory (endpoint_list, endpoint_get);
    sessions (session_list, login_storm_analysis); drift (drift_report,
    patch_status).
  • Write: endpoint_assign_profile (high, reversible — records an inverse
    reassign undo), endpoint_reboot (medium, no safe inverse, captures
    before-state).
• Two signature analyses — login-storm detection + slow login/boot ranking,
  and patch/config drift vs a fleet-majority (or explicit) baseline. Both accept
  injected records for offline analysis.
• Encrypted API key store (~/.endpoint-aiops/secrets.enc, Fernet + scrypt)
  — never plaintext on disk; legacy ENDPOINT_<TARGET>_APIKEY env fallback.
• CLI with an init onboarding wizard, secret management, and doctor.
• Bearer-auth REST connection layer over a generic endpoint-management REST
  API with teaching error translation (EndpointApiError).

Install

uv tool install endpoint-aiops
endpoint-aiops init
endpoint-aiops doctor

Caveats

• The REST paths are modelled generically (/endpoints, /sessions,
  /version, /endpoints/{id}/profile|reboot) and need live verification.
• Out of scope by design: enrollment/de-enrollment, image/OTA management, and
  any bulk destructive operation.