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