Quick Start¶
This walks through the visionstudio CLI and web dashboard. If you haven't installed it yet, see Installation first.
Start VisionStudio¶
This starts the Dolt database (if it isn't already running) and opens the dashboard in your browser at http://localhost:9400. Press Ctrl-C to stop.
Register a Repository¶
VisionStudio tracks initiatives against repositories it knows about. Register one with:
See visionstudio registry --help for the full set of registry commands.
Navigating the Dashboard¶
The sidebar has four sections: Initiatives (programs, standalone initiatives, and their RMIs), Repositories, Maturity, and Performance. See the Dashboard Guide for a full tour of each — what every panel shows, every clickable element, and what the status colors and icons mean.
Working with Initiatives from the CLI¶
The dashboard is a view over the same data the CLI manages. Common commands:
visionstudio initiative list
visionstudio rmi list --initiative <initiative-id>
visionstudio spec judge record <initiative-id> <spec-file> --score <1-5> --rationale "<why>" --model <model-id>
visionstudio roadmap import docs/specs/initiatives/<initiative-id>/ROADMAP.md # sync ROADMAP.md into the database
Every initiative has a spec workflow (which documents it needs — PBHQ Lite, AWS Working Backwards, Quick Fix, etc.), selected at creation and changeable later:
visionstudio workflow list # browse the catalog
visionstudio initiative create --workflow aws-two-way-door ... # select at creation
visionstudio initiative update <initiative-id> --workflow aws-one-way-door # switch later
See Spec Workflows for how required/optional/extra files and the workflow diagram follow the selection.
To keep the dashboard focused, programs and initiatives can each be hidden independently — hiding a program cascades to hide its initiatives too:
See Hiding Programs and Initiatives for exactly where this takes effect.
Run visionstudio --help for the full command list, or visionstudio <command> --help for any subcommand.
LLM-as-a-Judge Evaluation¶
Specs are evaluated against workflow rubrics (structured-evaluation/rubric.Rubric) — a 1–5 integer score, per-category findings with severity, and an overall pass/fail decision. See Specs & Evaluation for how results are recorded and displayed.
Agent Access (MCP)¶
Agent sessions can query initiatives, RMIs, and work assignments directly via the MCP stdio server: