AI behavioral intelligence layer

The only layer that sees
all your AI tools
at once.

Claude Code + Cursor + Windsurf + OpenCode — FORG connects them all without proxying a single request.

Full cross-tool visibility, live behavioral rules, and proof of outcomes for every session.

Metadata-firstOpen protocolNo proxy required
forg — bash
The problem

AI is invisible, uncontrolled, and unmeasured.

Your team is shipping with Claude Code, Cursor, and Codex. But you have no idea what's happening, why costs spike, or whether it's actually working.

01

No visibility

AI usage is spread across many tools with no unified view.

02

No control

Policy lives outside the workflow — impossible to enforce in real time.

03

No proof

Teams can't show what AI improved, what it cost, or what risks it introduced.

1
ObserveCapture AI work signals from every tool.
2
ClassifyDetect drift, cost overruns, and risk patterns.
3
ActNudge, alert, refocus, audit, or block.
4
MeasureShow adoption, safety, and business outcomes.
What FORG does

FORG turns AI activity into a managed workflow.

Users keep their tools. FORG adds the control plane — outside the request path, with no latency impact.

Observe → Classify → Act → Measure
How it works

Live in minutes, not months.

A lightweight adapter emits behavioral signals. FORG processes them server-side. No proxy. No latency. No prompt content.

01

Connect adapter

Claude Code, OpenCode, Codex, Gemini CLI, or any custom tool.

02

Emit signals

Sessions, model use, goal drift, cost, and outcomes — metadata only.

03

Run rules

Fast policy checks outside the LLM path. No latency in hot paths.

04

Show proof

Dashboards, audit trail, team summaries, and outcome tracking.

Step 1 — Connect tools

FORG starts with the tools people already use.

No workflow rewrite. Connect an adapter in one command and start observing real usage within minutes.

1 adapter to startOpen protocol
Adaptersconnected
Claude Codeofficial
OpenCodeofficial
Codex CLIcompatible
Gemini CLIcompatible
Custom CLIprotocol
event session.start
event model.call tokens=4,812 cost=$0.014
event goal.drift magnitude=0.72
event session.end quality=88

POST /engine/v1/signal
auth HMAC signed
latency < 0.5 ms
Step 2 — Capture signals

FORG watches metadata, behavior, and outcomes.

Default telemetry doesn't touch prompt or response content. Every signal is HMAC-signed and transmitted out-of-band at under 0.5ms overhead.

Metadata-firstOut-of-band
Steps 3–4 — Rules & dashboards

Rules make AI governance practical.

Alert first. Enforce harder only when needed. Operators get one clear dashboard — who's using AI, where it's helping, where work drifts.

Per-user and per-project attribution
Goal drift detection with refocus nudges
Cost anomaly alerts to Slack
Immutable audit trail for compliance
Rulesactive
Goal drift > 0.70refocus
Budget at 80%notify
Unknown workspacereview
Revoked licensekill-switch
Team dashboardhealthy
82%
active
18%
less drift
91%
aligned
Who benefits

Value for every layer of the org.

Individual users

Stay focused

Know when AI work leaves the goal.

Save time

Catch drift before a session goes too far.

Show results

End sessions with clearer outcomes.

Team leads

See adoption

Know which tools and workflows are active.

Compare patterns

Find where AI is helping or hurting.

Improve safely

Coach workflows without reading private work.

Security & platform

Reduce blind spots

Unknown tools and risky patterns become visible.

Apply policy

Rules work where AI usage actually happens.

Keep evidence

Audit trails support internal review.

Trust model

Designed for trust.

Useful visibility without making people feel watched. FORG is built so operators can govern AI usage without ever reading private work.

“Still working on the billing fix? You're editing unrelated auth files.”

1
No prompt content by defaultTelemetry is metadata-first. No prompts, no completions.
2
No LLM in real-time rulesFast, cheap, predictable policy checks — zero inference latency.
3
Data split is clearLicense/identity on D1. Behavioral data in Supabase. Never mixed.
Real-time analytics

Live token usage across your entire org.

Drill down by team, project, user, or model. Set alerts for anomalies. Export to your data warehouse.

Per-user and per-project cost attribution
Model cost breakdown by session
Anomaly detection with Slack alerts
CSV / BigQuery / Snowflake export
Daily Cost Trend↓ 12.4% MoM
62%
Claude
23%
GPT-4o
15%
Other
Pricing

Simple, transparent pricing.

Solo at $19/mo. Scale as your team grows.

Solo
$19/mo

For individual engineers who want full visibility into their AI usage.

1 seat
5 adapter slots
30-day history
$15/mo billed annually
Get started
Professional
$39/mo

Unlimited adapters, 12-month history, and AI insights.

1 seat
Unlimited adapters
12-month history
$31/mo billed annually
Get started
Team
$199/mo

Org-wide visibility, team analytics, and API access.

Up to 20 seats
+$12 per extra seat
Team analytics
API access
Get started
Compare all plans
Get started today

Adopt AI faster because you
can finally see and govern it.

FORG makes AI work visible, guided, and measurable. Starting at $19/month.

Get FORG Read the docs