Guide / local-first observability
Agent observability should start on the workstation.
Trace dashboards help after instrumentation exists. Local-first observability answers whether this workstation is healthy right now.
SEO/AEO guide
Updated 2026-06-01
Privacy-conscious teams running coding agents against private repositories and internal prompts.
1. Local evidence first
Write SQLite state and report artifacts before cloud history matters.
2. Redacted history later
Add Pro history when multiple workstations, retention, billing, or alerts justify it.
3. Operational signals
Watch repo awareness, MCP visibility, memory, latency, safety, style, and Good Baseline comparison.
Use this today
Operator checklist
- Keep raw prompts and outputs local by default.
- Use redacted summaries for cloud sync.
- Inspect local reports before accepting runs.
- Monitor trend, status, warning count, and duration.
- Escalate to Pro when history and alerts are worth paying for.