Payment authorization layer for AI agents with enforced spending limits
AI runs where you own it
Stage
Not given
Not given
Founded
Not given
Not given
Based in
US
Not given
Pricing model
Paid
Freemium
Pricing
3¢ per successful transaction (flat fee, no volume tiers). For transactions under $1, the fee is 3% with a 0.5¢ minimum. Connecting a wallet and setting spending limits is free for users; only merchants pay transaction fees.
RailCall local runtime: free. RailHub Team: $5 per seat per month. Cloud agents: $5 per trained agent per month. Advanced thinking (overflow): metered against a budget you set, with 500 flows per month included. Storage on hub: $0.25 per GB-month with 5 GB free.
Key features
Hard spend limits with atomic enforcement preventing overspending to the cent
Cryptographic spending mandates that agents cannot override
Merchant allowlists restricting agent payments to approved vendors
Auto-expiration and instant revocation of spending mandates
Settlement via Stripe (190+ countries, 135+ currencies) or USDC on Base
Zero credential exposure—agents never see card numbers or private keys
Automatic endpoint discovery on Delegare Market, agentic.market, and MPPScan
Studio visual workflow builder with 1,000+ nodes
CLI and terminal-first interface
Ed25519-signed tamper-evident receipts
Human approval airlock for all actions
Autonomous agents that run 24/7
Capacity routing across organization machines
MCP integration with 33 tools
Multi-level human approval workflows
Offline receipt verification
Company-wide stop button
What makes it different
Only 3¢ per transaction with no volume tiers or minimum transaction amounts, compared to ~$1.03 for Stripe base fees on $25 transactions
Merchants integrate in 5 minutes with just 5 lines of code
Agents discover and pay endpoints without manual registration or configuration
Transactions under $1 cost only 3% (minimum 0.5¢) making micropayments commercially viable
Entire local runtime free with MIT license, no login required
Cryptographically signed receipts verifiable offline