Put a policy-driven human gate in front of agent actions that move money, touch sensitive data, or create regulatory exposure.
Pactline gives teams one approval layer for high-risk AI workflows: real-time policy evaluation, human review when needed, and audit-grade evidence you can export when regulators, customers, or security teams ask what happened.
Agents can send external communications, move money, access patient records, or trigger workflow changes in seconds. The higher the autonomy, the higher the need for explicit policy and human control.
When security, compliance, or customers ask who approved an AI action and why, many teams are left with Slack threads, email chains, and incomplete logs instead of exportable evidence.
Building approval logic, queues, notifications, audit trails, and receipts in-house delays launches by months. Pactline gives teams the control point and evidence layer upfront.
Every high-risk action your agent wants to take is checked against your policies in real time — auto-approve when it is safe, block when it is not, and route to a human when it needs judgment.
Approvers review the action with full context — what the agent wants to do, why the policy routed it, and what the risk is — and approve or deny in one click from the dashboard, email, or webhook.
Every decision — human or automated — writes a hash-chained audit entry and a portable receipt you can export when regulators, customers, or security teams ask what happened.
Watch a real approval flow — from a high-risk agent action to a signed, exportable evidence record.
Should MedRecord AI be allowed to read and analyze patient medical records?
{
"@context": "https://www.w3.org/2018/credentials/v1",
"type": "ConsentCredential",
"issuer": "did:pactline:consent-authority",
"agent": "MedRecord AI",
"action": "Read and analyze patient medical records",
"riskLevel": "high",
"hash": "sha256:7b0cbc...e4f2a1",
"issuedAt": ""
}
Define the rules that decide which agent actions are safe to auto-approve, which need a human, and which should be blocked. First-match evaluation with clear conditions — no black boxes.
A focused dashboard for the humans in the loop. See each pending action in context, understand which policy routed it, and approve or deny with one click — from web, email, or webhook.
Every decision writes a hash-chained audit entry and a portable receipt. Export the full evidence bundle when auditors, regulators, or customers ask for proof of oversight.
Pain: AI copilots and agents trigger payments, KYC checks, credit decisions, and account changes autonomously. When something goes wrong or an auditor asks "who approved this?", teams cannot produce a clean trail.
Approval: Policies auto-approve low-risk operations, block prohibited ones, and route sensitive actions — high-value transfers, external disclosures, new-account openings — to a human with the exact context needed to decide in seconds.
Evidence: Every action produces a hash-chained audit entry and a signed receipt. Export the evidence bundle for internal audit, external regulators, or customer trust reviews.
See it in action →Pain: Clinical assistants and workflow agents access patient records, order labs, and generate care summaries. Every touch is a HIPAA event, and manual oversight processes cannot keep up.
Approval: Policies auto-approve routine read operations, block prohibited PHI access, and require human approval — from a clinician, DPO, or compliance officer — for anything that changes care or leaves the system.
Evidence: Every access is recorded and receipt-signed, ready to be included in HIPAA audits or attached to a specific incident review.
See it in action →Pain: Automated underwriting, claims triage, and adjudication move fast — but regulators, ombudsmen, and reinsurers expect a defensible human-in-the-loop for the decisions that matter.
Approval: Policy-driven routing sends high-value claims, denials, and cross-jurisdiction decisions to a human reviewer with the full case context. Low-risk decisions still auto-approve.
Evidence: Each decision produces exportable evidence — action, actor, policy, risk level, and outcome — attached to a specific claim ID for retrieval later.
See it in action →Pain: Multiple teams ship agents on top of the same LLM platform. Security, legal, and risk teams need one control plane instead of chasing individual pipelines.
Approval: A single API and policy layer sits in front of every high-risk agent action across teams — with role-based access, per-agent policies, and centralized approver queues.
Evidence: Organization-wide audit trail with filtering by team, agent, and outcome. One export path for security reviews and internal governance.
See it in action →Pactline is engineered to help teams meet human-oversight and evidence requirements in these frameworks. Formal certifications are on the roadmap — see our trust posture.
Wrap any function your agent calls with Pactline.gate(). Every invocation is checked against your policies — auto-approved when safe, routed to a human when it needs judgment, and blocked when the policy says so. Every decision produces an audit-grade receipt.
pip install pactlinePactline(api_key=...)@pact.gate(agent_id=...)call() as normalfrom pactline import Pactline
pact = Pactline(api_key="pk_live_...")
# Any function your agent calls can be gated.
# Pactline evaluates policies, routes to a human if needed,
# and writes an audit-grade receipt.
@pact.gate(agent_id="agt_...", risk_level="high")
def transfer_funds(from_account, to_account, amount):
return banking_api.transfer(from_account, to_account, amount)
# Normal call — approval + evidence happen automatically.
transfer_funds("acct_a", "acct_b", "15000")import { Pactline } from '@pactline/sdk';
const pact = new Pactline({ apiKey: 'pk_live_...' });
// Wrap the function your agent calls.
// Pactline evaluates policies, routes to a human if needed,
// and writes an audit-grade receipt.
const gatedTransfer = pact.gate(transferFunds, {
agentId: 'agt_...',
riskLevel: 'high',
description: 'Move funds between accounts',
});
// Normal call — approval + evidence happen automatically.
await gatedTransfer('acct_a', 'acct_b', '15000');package main
import pactline "github.com/pactline/pactline-go"
func main() {
pact := pactline.NewClient("pk_live_...")
// Create a consent request and wait for a decision.
result, err := pact.Consent.Create(pactline.ConsentRequest{
AgentID: "agt_...",
ActionTitle: "Transfer $15,000 to acct_b",
RiskLevel: "high",
})
if err != nil { return }
if result.Status == "approved" || result.Status == "auto_approved" {
transferFunds("acct_a", "acct_b", "15000")
}
}# Create a consent request for a high-risk agent action
curl -X POST https://pactline.io/api/consent \
-H "X-API-Key: pk_live_..." \
-H "X-Agent-Id: agt_..." \
-H "Content-Type: application/json" \
-d '{
"actionTitle": "Transfer $15,000 to acct_b",
"actionDescription": "Move funds between customer accounts",
"riskLevel": "high",
"context": { "amount": 15000, "currency": "USD" }
}'
# Response
{
"consentRequestId": "cr_...",
"status": "pending",
"policyDecision": "require_human",
"policyId": "pol_..."
}Start free. Every plan includes approval, human review, audit trail, and exportable evidence.
Enough to build a real prototype with approval + audit.
Production oversight for one product line.
Multi-team oversight with SLA and evidence export.
Custom deployment, retention, and evidence controls.
EU AI Act
Human oversight in front of high-risk AI decisions
GDPR
Hash-chained evidence you can export on demand
HIPAA
Approval and audit for clinical AI workflows
Approval, oversight, and exportable evidence for the AI actions that matter most. Free to start.
Protect Your First ActionHave questions about Pactline? We'd love to hear from you.