Per-Agent Coding CLIs and Clearer Agent Alerts

Each agent can now save its own coding CLI and model, credential setup covers more provider combinations, and execution alerts are easier to understand.

This day made agent execution feel more intentional. Instead of treating the coding toolchain as a global afterthought, Kindship started letting each agent carry its own execution setup, while agent alerts became a little more readable and outcome-focused.

Inner Loop Settings Became Per-Agent

Every agent can now carry its own coding CLI and model preference instead of inheriting a one-size-fits-all default.

  • You can save a coding CLI and model on each agent
  • Those saved settings now belong to the agent profile itself, which makes specialized agents easier to tune for the kind of work they do
  • Credential management covers more real provider combinations, including better OpenCode alignment and broader multi-CLI setup support

The Coding Tooling Mix Broadened

Kindship also became more comfortable with a wider coding-agent fleet instead of leaning so heavily on a single default.

  • Gemini CLI joined the broader supported tool mix
  • Inner-loop saves and credential picking became less brittle, so changing execution setup feels quicker and less error-prone
  • Agent execution is better prepared for mixed-CLI workflows, which matters when different agents have different strengths

Execution Alerts Got More Useful

Agent notifications are starting to say what happened, not just that something happened.

  • Execution notifications now lean more toward results than raw activity
  • Email delivery became more dependable in practice, with clearer formatting for longer messages
  • Heartbeat controls also picked up a simple pause/resume flow, which gives recurring agents a cleaner way to be temporarily quiet without deeper reconfiguration