Agents · Runners
The AI workforce: domain-specialized agents that accumulate institutional knowledge over time.
Runners are the agents that operate within Runlines, skill- and process-optimized for their domain. Each accrues institutional knowledge about the credit union it serves and delivers roughly 2 to 3 FTEs of annual capacity, with room to scale.
Its facets:
- Prime Runners: the durable, long-lived Runner for a domain at a specific CU (documented under Runlines).
- The Tower: where a Runner's work is watched and directed (documented under Interfaces & Apps).
- Attention: how a Runner decides what to work on next (below).
- Trust Tiers & Lifecycle: how a Runner earns autonomy over time (below).
Attention
Attention is the queue and prioritization primitive that decides what a Runner works on next (the @runline-ai/attention package): the attention queue, prioritization and reactive triggers, all tied into Circuit events.
Trust Tiers & Lifecycle
Runners graduate through trust tiers: training_wheels → supervised → semi_autonomous → autonomous. A Runner earns autonomy through demonstrated reliability, and the Grid enforces its limits regardless of tier.