Your task
Independent AI workflow planner
Choose the right
workflow.
Match the job to a workflow, define what counts as evidence, and set the boundary for uncertainty and private data.
Scope
Set the standard
Privacy
Control the export
Claims
Separate what you know
Evidence
Build source lineage
Generated route
Research with traceable evidence
Compare routes
Different jobs need different evidence.
Take the route with you.
The exported brief records the workflow, evidence boundary, privacy choices, and unresolved uncertainty.
A local plan separates routing from execution
Read the explanation
For an illustrative task with one named high confidence source, a three source target, one fact and no open questions, readiness is thirty five plus twelve plus fifteen plus ten plus five, or seventy seven percent. Supplying three named sources raises the evidence term to thirty five and the total to one hundred. Bars share five pixels per model point. Source confidence is user declared, not authentication or a successful research result. The saved router offers five workflow categories, each with five stages. Bars share ninety pixels per count. Task text receives one point for each keyword substring present, and ties retain source ordering. A manual route stays selected after task edits. This local keyword planner does not call a language model, execute the workflow, retrieve sources or certify that any stage was completed. The saved example has three named sources, but only two are included in export. Bars share one hundred fifty pixels per row. Readiness counts named sources even when excluded. Local-only removes task text, fact and assumption notes, open question notes and source URLs from export, while source titles and metadata still leave the workspace. Pattern redaction has limits. Retain stores local state only when enabled; export preserves source lineage without claiming external verification.