Control layer for Agentic Commerce
Agentr lets AI agents buy safely.
Give every AI agent a budget, a card, an approval flow and an audit trail. Connect from ChatGPT, Claude, WhatsApp or MCP. Finance keeps the controls.
PCI SOC 2Stripe IssuingMCP-readyAgentic Commerce
The control loop
Every purchase your agent attempts passes through Agentr.
Step 1
Agent requests
ChatGPT, Claude, WhatsApp or MCP call the purchase API.
Step 2
Policy enforced
Budget, per-tx, merchants, categories, all server-side.
Step 3
Human approves
Inbox-style center. Slack / WhatsApp ping. One click.
Step 4
Card released
Stripe Issuing virtual card with the exact constraints.
Step 5
Audit recorded
Who, what, when, why. SOC 2-grade trail.
Spend with control
Agents and people request purchases. Policy decides. A human approves. A scoped virtual card pays. Every step audited.
Start controlling spendSell to agents and humans
Publish a checkout link. A person pays with a card, or another AI agent reads the manifest and pays within its policy.
Create a checkout One endpoint, every agent
POST /api/v1/purchase-requestscurl -X POST https://agentr.com.br/api/v1/purchase-requests \
-H "Authorization: Bearer agk_<your_agent_key>" \
-H "Content-Type: application/json" \
-d '{
"vendor": "OpenAI",
"amount": 200,
"rationale": "Upgrade to Team plan, unblock 8 engineers"
}'
# → { "id": "rec_...", "status": "pending",
# "next_action": "Awaiting human approval in Agentr." }Built for SaaS buying
AI agents that research, recommend and buy SaaS, with human approval and controlled virtual cards.
Marketing Agent
Spends on Meta, Google and LinkedIn ads, within CAC and channel limits.
Operations Agent
Renews SaaS, catches duplicate subscriptions, negotiates plan downgrades.
Procurement Agent
Books travel, supplies and vendor contracts, every approval auditable.