Every Job Knows Its Purpose
Company objectives, project targets, and task goals flow down into every expert agent execution. Agents work with strategic context, not just instructions.
Company objectives, project targets, and task goals flow down into every expert agent execution. Agents work with strategic context, not just instructions.
Q2 GOAL
Reduce auth
latency 40%
Migrate to JWT
auth token refactor
Redis Session Cache
add redis adapter
Organization-wide objectives. Set once, inherited by all projects and tasks. Examples: ship v2 platform, achieve SOC 2 compliance.
Repository or initiative-scoped targets. Link to a specific project and its jobs. Examples: migrate auth service, reduce p99 latency.
Atomic work-item goals. Attached directly to a job. Examples: add MFA endpoint, fix null pointer in checkout flow.
When a job has a goal, the full ancestry chain — from company down to task — is assembled and injected into the expert agent's context before execution. Agents understand why they are working, not just what to do.
getAncestry() walks the goal tree up to 10 levels deep, returning the root-to-leaf path as structured Markdown.
Goal context is injected before guardrails (priority 1.0), ensuring strategic intent frames every constraint check.
Each job carries an optional goalId. The scheduler, manual triggers, and webhook handlers all support goal association.
Goals are not documentation. They are live context that shapes every decision your expert agents make.
Request Early Access