Private Beta

Let humans approve, edit, or take over risky AI agent actions

Add secure human review checkpoints to AI agents in minutes. Pause risky actions, notify a reviewer, collect a decision, and resume safely with an audit trail.

Built for teams using LangGraph, CrewAI, Vercel AI SDK, n8n, and custom agents.

How It Works

Human control for autonomous AI

Forsig creates a human checkpoint when your AI agent is about to do something risky. Simple, secure, auditable.

01

Secure Review Checkpoints

Create signed, time-limited review pages for human oversight. Every decision is auditable.

02

30-Minute Integration

Simple API and TypeScript SDK. Add human intervention to any agent workflow quickly.

03

Framework Agnostic

Works with LangGraph, CrewAI, Vercel AI SDK, OpenAI Agents, n8n, Make, Zapier, and custom agents.

04

Full Audit Trail

Track who reviewed, what they decided, when, and why.

05

Multi-Channel Ready

Start with web and email. Slack and other review channels can follow based on demand.

06

Safe Timeout Defaults

If no human responds, risky actions should fail closed, not open.

Interactive Demo

Experience the review page

This is what reviewers see when an AI agent requests human intervention.

!

High Risk Intervention

Expires in 28 minutes

Pending

Review refund for customer #123

The agent wants to issue a $500 refund to a VIP customer.

Agent Refund Agent v1.2.0

refund-review-flow / step: approval

Risk Level High

Refund exceeds $250 policy limit

Proposed Action
{
  "type": "refund",
  "amount": 500,
  "currency": "USD",
  "refund_method": "original_payment_method",
  "customer_id": "cus_123"
}
Context

Customer Tier: VIP

Order Value: $1,200

Refund Reason: Product failed twice

Your Decision

Integrate in minutes, not days

A simple SDK that works with any agent framework. Create human intervention checkpoints with a few lines of code.

1 Agent calls forsig.intervene()

2 Human receives notification and reviews

3 Decision returns to your agent

agent.ts
import { Forsig } from "forsig";

const forsig = new Forsig({
  apiKey: process.env.FORSIG_API_KEY
});

const decision = await forsig.intervene({
  agent: { id: "refund-agent", name: "Refund Agent" },
  task: { title: "Review refund for customer #123" },
  risk: { type: "refund_over_limit", level: "high" },
  proposedAction: { type: "refund", amount: 500 },
  waitForDecision: true,
  timeout: "30m"
});
Pricing

Simple private beta pricing

Free during private beta. Paid tiers are planned for production teams.

Free

For testing and small projects

$0
  • 100 interventions/month
  • Email notifications
  • Web review page
  • Basic audit log
  • 1 workspace
Get Early Access

Starter

For growing teams and agencies

$49/month
  • 1,000 interventions/month
  • Email + Slack notifications
  • Custom branding
  • Full audit log
  • Priority support

Pro

For production workloads

$199/month
  • 10,000 interventions/month
  • All notification channels
  • Advanced roles
  • Unlimited workspaces
  • Dedicated support

Join the private beta

Be among the first to add human control to your AI agents.

No spam. We will only email you about Forsig updates.

Talk to the founder

Share what you are building, where your agents need human review, or what would make Forsig useful for you.