FORG gives engineering managers real-time visibility, hard budget caps, and a tamper-proof audit trail — for every AI tool across every squad.
The problem
Claude Opus quietly racked up $25K in a single sprint. Nobody noticed until the AWS bill arrived. No alerts, no visibility, no one accountable.
Finance asks for an AI spend breakdown. You open six dashboards, export three CSVs, and guess. It takes a day — and the numbers still don't match.
Audit season hits. Lawyers ask for a log of every AI session and which data left the org. You have nothing. GDPR risk is real and growing.
Setup
Drop one line into your Claude Code settings. Works with Cursor and VS Code too. Takes 5 minutes.
# In ~/.claude/settings.json
{
"proxy": "https://forg.pro/engine/v1",
"apiKeyHelper": "forg auth token"
}Define hard caps per team, per model, and per engineer in a single config file.
{
"teams": {
"platform": { "monthly_usd": 4000 },
"growth": { "monthly_usd": 2500 },
"infra": { "monthly_usd": 1500 }
}
}Real-time spend, budget utilisation, and flagged sessions — no extra tooling needed.
Dashboard preview
What's included
Hard stops at team and individual level. When the cap is hit, the request is blocked — not just logged. No exceptions unless you explicitly allow them.
Works with Claude Code, Cursor, and VS Code out of the box. One config change per developer. No proxies to self-host, no infra to maintain.
Every session, every token, every model — logged immutably. SOC 2 ready. Export to your SIEM or compliance tool in one click.
FORG cut our AI overspend by around 60% in the first month. I finally have visibility across 8 squads without pestering anyone.
Most engineering managers see their first org-level insight within 5 minutes of install. No infra required.