See every model call. Own what it costs.

FORG meters every request on-device and rolls it up by repo, seat, and model — so the cost of AI engineering stops being a mystery line on the invoice.

$ forg session --today
requests   312  metered on-device
tokens     84.1k in · 31.2k out
cost       $3.82  across 3 models
top        claude-sonnet-4-6 · 51%
saved      $0.94  cache + downgrade
 

Per-session cost

Every model call is timestamped and priced on-device. No estimates — actual token counts from the response.

Model breakdown

See spend by model, repo, and seat. Know which team, which tool, and which model is driving the bill.

Export

Pull a finance-ready CSV or evidence pack for any window and any scope in a single click.

Replay any session as a cost graph

Every session is a graph of tool calls, token counts, and model decisions. Scroll back through any window and see exactly what drove the spend — no guessing, no sampling.

Session log · today