Agent Payments: Lobster.cash + Treeship

Lobster.cash (by Crossmint) gives agents a payment layer: wallet provisioning, transaction signing, settlement. Treeship adds the verification layer: every transaction gets a signed receipt.

Together, they solve both halves of the agent payment problem. Lobster.cash answers "how does the agent pay?" Treeship answers "can anyone prove the agent was authorized to pay?"

Separation of concerns

The two systems own different things:

This separation matters. Payment infrastructure and trust infrastructure have different threat models. Combining them creates a single point of failure. Keeping them separate means each system can be audited, upgraded, and secured independently.

The flow

1. Agent gets human approval

treeship attest approval \
  --approver human://ops \
  --description "approve $49.99 USDC for API credits"

2. Agent calls Lobster.cash to execute payment

The agent calls the Lobster.cash API to send USDC. This is a standard API call, nothing Treeship-specific.

3. Agent attests the result

Option A: Wrap the API call directly.

treeship wrap \
  --approval-nonce abc123def456 \
  -- curl -X POST https://api.lobster.cash/v1/payments \
    -H "Authorization: Bearer $LOBSTER_TOKEN" \
    -d '{"amount": "49.99", "currency": "USDC", "recipient": "0x..."}'

Option B: Attest the result after the fact.

treeship attest action \
  --actor agent://payments \
  --action lobster.payment.create \
  --approval-nonce abc123def456 \
  --meta '{"tx_id": "7f2a...", "amount_usdc": 49.99, "recipient": "0x..."}'

Both produce a signed receipt that binds the payment to the approval.

3b. Spend-limit proof (automatic)

When a payment is attested, Treeship runs the spend-limit-checker ZK circuit. This produces a zero-knowledge proof that the amount falls within the approved spend limit, without revealing the exact threshold to third-party verifiers. The proof is embedded in the receipt automatically.

4. The receipt chain is complete

The chain now proves:

human://ops approved "$49.99 USDC for API credits"
  -> agent://payments intended lobster.payment.create
    -> Lobster.cash executed tx 7f2a...
      -> receipt signed and chained

5. Anyone can verify

treeship verify --full

The receipt travels with the transaction. Verification happens offline, client-side. No callback to Treeship or Lobster.cash required.

Why this pairing works

Lobster.cash solves the hard problem of agent wallets: provisioning, key management, transaction signing, settlement finality. Building this from scratch is months of work and a significant security surface.

Treeship solves the hard problem of provable authorization: who approved, what was approved, did the action match the approval. Building this from scratch is also months of work and a different security surface.

An agent using both gets wallet infrastructure and trust infrastructure without building either. The receipts from Treeship and the transaction confirmations from Lobster.cash form a complete audit trail from human intent to settled payment.

Integration docs and Verifiable Intent

For a step-by-step setup guide, see the Lobster Cash integration page.

Lobster Cash payments are also compatible with Verifiable Intent credentials. When both are configured, Treeship produces the agent_attestation field for VI Layer 3 credentials, giving the payment a standards-compliant proof chain on top of the Treeship receipt chain.