Single-project briefs
Ask about any PJM project by ID. Get identity, cost context, risk, cluster partners, queue position, and an analyst takeaway grounded in real filings.
LIVE
The AI analyst for the US power grid.
What we're building
An analyst that reads every filing, so you don't have to.
Power market data is fragmented across queue reports, cluster studies, cost allocation files, regulatory dockets, and ISO stakeholder meetings. Developers and investors still piece it together by hand, week after week.
Gridsurf fuses two layers. A normalized data layer that turns messy PJM source files into structure a model can actually reason over - and an intelligence layer, GridAgent, built to answer the questions a seasoned analyst asks every day, with citations back to the file that settled the argument.
01
02
Cluster cascade modeling
See every co-allocated project on every shared network upgrade. Understand who inherits the cost when a partner drops.
03
Citations on every claim
Each answer cites the files it read. No hidden prompts. No training-data guesses. Click a citation, read the source.
04
Sandboxed execution
Python, grep, and pandas run inside isolated per-user sandboxes. Real code against real data, inside real boundaries.
Who we are
Builders who have read too many cluster studies.
Gridsurf is a small team with experience in power markets, grid engineering, and AI systems. We started this after watching analysts, developers, and investors spend weeks reading the same queue reports and cluster studies by hand - and noticing that the constraint was never the data. It was the work of understanding it.
We're building the tool we wished existed in the room when the real decisions got made.
Power markets
Interconnection, market dynamics, and project risk across PJM, SPP, MISO, and ERCOT.
Grid engineering
Utility-scale solar and storage. Prior work on tooling for querying and analyzing grid data.
AI systems
Retrieval pipelines, tool-using agents, and production systems that stay honest under load.