Agent console
Research and advisory agents
Two models share the work: Muse Spark (muse-spark-1.3) gathers, extracts and plans, and Muse Wonder (muse-wonder-1.0) writes each verdict and build plan, falling back to Spark if it is unavailable. Web search runs through Muse search grounding with Tavily as backup, SAM 3.1 measures each enriched site from satellite imagery, and Muse Image renders concepts. Every run is logged with its sources, model and token use.
Discovery
Web search over operator newsrooms, utility filings and trade press, then structured extraction. Dedupes by name and country; never overwrites curated entries.
Enrichment
Per-site research into power price, grid carbon and headroom, interconnection queues, labor, land, water stress, population and sentiment, written into the region profile with sources. Then SAM 3.1 looks at the site from above and measures its open land, buildings, water and solar, which replace the priors in every estimate.
Advisor
Runs the energy-mix search against the viability floors, fixes the verdict deterministically, then asks the model for the rationale, conditions, risks and a brief per audience. Run it from any site page.
Planner
A conversation over the same engine, on the simulation page. Describe a facility and a place; the model looks the place up, estimates it, compares power mixes and hands back a plan you can build in the 3D world.