The payment gate for agent work.
Payjent lets your agent ask a person to approve and pay for premium work — once, at the right moment — then resumes the exact action with a receipt-backed ledger.
Quote, approve, grant, capture.
Agent quotes the work
Your agent tells Payjent what action it wants to take and at what price. Payjent binds the quote to the exact request and execution envelope.
Human approves once
Payjent presents an approval card over the chat surface, email, Slack, or webhook. A person approves a specific amount for a specific request.
Single-use access issued
Approval creates a request-bound, single-use authorization. It only unlocks the original action — never replayed or re-scoped.
Action resumes, receipt logged
The agent resumes the exact action it quoted. Payjent records the checkpoint and ledger trail; it does not execute downstream pay.sh.
The wedge is where money meets intent.
The checkpoint agents never had.
Other agent payment stories start with a budget and assume the agent may spend. Payjent treats payment as a deliberate human checkpoint bound to one action and one amount.
One registration. Many premium actions.
Register an agent once, point it at the machine-readable setup guide, and route paid actions through one approval and receipt system instead of scattering vendor secrets.
Receipts that survive an audit.
Each spend record keeps the prompt, quote, human decision, payment session, capture, and fulfillment evidence together so finance can see why money moved.
Designed so nothing surprising ever leaves your account.
- 01
Request-bound approvals
Approval is tied to the exact action envelope. Re-quoting requires re-approval.
- 02
Single-use by default
Access is minted with one use and short TTLs. Replay attempts fail closed.
- 03
Per-agent caps & policies
Daily caps, vendor blocks, category policies, and risk checks communicate or enforce conservative limits.
- 04
Credentials shown once
Bot keys are revealed exactly once at registration, then hidden from dashboards, logs, and support flows.
Two steps. No SDK snippets on the landing page.
Point your agent at the setup contract
Tell Claude, Cursor, Copilot, or your own agent to integrate with Payjent and give it the canonical setup path. The agent reads the contract and wires itself up.
- Works with any agent framework
- Machine-readable self-setup endpoint
- No raw tokens on public pages
Register the agent in your dashboard
Mint a credential bound to one agent identity. Drop it into your agent secret store. From then on, paid actions flow through the human approval gate and ledger.
Same problem, different shape.
| Payjent | Stripe Agent SDK | x402 / buildx402 | pay.sh | DIY | |
|---|---|---|---|---|---|
| Human approval at the moment of payment | ● native | ○ no | ○ no | ◐ partial | ○ build it |
| Request-bound, single-use approvals | ● native | ○ no | ◐ partial | ○ no | ○ build it |
| Reason-backed ledger (prompt → receipt) | ● native | ○ no | ○ no | ○ no | ○ build it |
| Works across Stripe, x402, and custom rails | ● rail-aware | ◐ one rail | ◐ one rail | ◐ one rail | ◐ maybe |
| Per-agent caps, vendor blocks, daily limits | ● native | ◐ partial | ○ no | ◐ partial | ○ build it |
※ Stripe, x402, pay.sh and others ship useful primitives — Payjent is the human checkpoint and receipt layer around paid agent work. Payjent does not execute downstream pay.sh.
Register an agent.
Two minutes. One credential. Shown exactly once. Drop it into your agent secret store and Payjent will gate the next paid action.