Give agents the code relationships they need—bounded to the task.

From discovery to review

Managers and workers can use focused symbols, file slices, surrounding context, call traces, dependencies, impact maps, test maps, hotspots, coverage, churn and task-oriented bundles. Query choice follows the workflow stage instead of defaulting to a full-tree read.

Token economy is measured, not assumed

A smaller context packet is not automatically a useful one. AIWorkHub records tool usage, requested and delivered bytes, truncation and acknowledged receipts so Source Graph benefit can be compared against direct file-reading baselines.

Configurable repository coverage

The index supports configurable code, data and documentation families, including structured formats such as JSON and XML. Repository settings decide which families are enabled; readiness and index revision remain observable.

Evidence labels and freshness

Extracted, inferred and ambiguous relationships stay distinguishable. Code tasks can require a non-empty, fresh Source Graph packet before work begins, preventing a stale or empty index from silently passing as context.