Agentcard is a debit card for ai agents. Give your OpenClaw or Claude an safe way to pay for things online,
This is the 2nd launch from Agentcard. View more
Buy by Agentcard
Launched this week
you can now buy Doordash from Claude. connect Agentcard to your Openclaw, Claude or whatever you use, ask for what you want and we handle the rest. it feels magic.
install from agentcard.sh and make your first order for free!
Free Options
Launch Team
Agentcard
Genuinely curious how you're handling the fraud surface here — when an agent can place an order, what stops a prompt-injection or a spoofed merchant from turning "order dinner" into a drained card? Is there a spend cap or a human-confirm step, or does the user set the guardrails up front?
Agentcard
@peterdigitalis great question! cards are one time and created on the spot to complete a single order. they get destroyed after use :)
@keyserfaty Nice, single-use is a clean containment design. Does it also lock to a set merchant + amount at creation? That's the gap single-use alone doesn't close if the agent gets fed a bad instruction.
Agentcard
Ordering food is convienient, yet I would hesitate before letting an agent make higher-value purchases. Curious how you think about building confidence over time.
Agentcard
Bluerails Discovery
How does the agent actually make the payment? It uses chrometools to select the product and navigate to checkout?
Agentcard
@ashwin_kumar46 we don't use chrome to make this, it would spend too many tokens. you should try it! for payment we create an Agentcard and pay :)
great product guys! Hope this becomes the default way to shop in the agent economy era.
What are the ways to fund your card? do you guys support x402?
Agentcard
@arag_agrawal we don't. you fund with a credit/debit card.
Zaro
Cool product! Curious how guardrails will work
Agentcard
Starting with DoorDash is a smart wedge, a bounded low-stakes purchase to prove the buy flow before agents touch anything bigger. For someone running several agents, can I issue a separate card per agent or per workflow so limits and spend are scoped independently, and does each transaction carry back enough metadata to reconcile which agent/run made the purchase? Trying to picture how the accounting maps back when it's not just one Claude ordering dinner.