Interline

Interline

A neutral, non-custodial payment router for AI agents: it discovers which payment rails a paid endpoint accepts (x402 on Base/EVM, Solana, MPP), then pays across them using the caller's own wallet key. Three tools — discover_payment_rails, pay_for_resource, payment_history.

Category
Visit Server

README

Interline — agent-to-agent payment router

An agent-to-agent payment router: one agent does work, another agent pays for it, no human keys a card. Interline is a neutral, non-custodial router over the fragmented agent-payment-rail stack — "OpenRouter for agent payments." Built on x402 (HTTP-402 micropayments, USDC), aggregator-shaped so adding a rail is one adapter, not a rewrite.

Today it routes payments across three live railsx402 USDC on EVM (Base Sepolia) and on Solana (devnet), plus MPP on Tempo (Moderato testnet) — all behind one Paywall.gate() call. Real on-chain agent-to-agent settles are confirmed on all three. It ships an MCP router so agents can discover + pay endpoints, and includes a Google-AP2 inbound adapter — a signed AP2 mandate (verified, constraint-checked + freshness-gated) settles non-custodially on the rails.

Run the demo (no wallet, no faucet, no risk)

python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
python3 run_demo.py

Expected: DEMO PASS — agent paid agent, settlement receipt issued.

What just happened (the x402 "exact-evm" loop):

buyer GET /work
  -> seller 402 + PaymentRequirements {amount, asset=USDC, payTo, network}
  -> buyer signs an EIP-3009 transferWithAuthorization (gasless USDC auth)
  -> buyer retries with X-PAYMENT header (base64 signed payload)
  -> facilitator VERIFIES the signature (real signer-recovery) + checks policy
  -> facilitator SETTLES (mock: fake tx hash; live: on-chain USDC transfer)
  -> seller 200 + work product + X-PAYMENT-RESPONSE receipt (tx hash)

The facilitator's verify step is real cryptography even in mock mode — it recovers the EIP-712 signer and matches it to the authorization. Only the on-chain broadcast is mocked. So a passing demo proves the buyer's signing is protocol-correct against real USDC.

Files

file role
router/paywall.py the reusable Paywall primitive — gate any endpoint behind x402 in one .gate() call
router/ledger.py settlement receipt-ledger → logs/agent_payment_settlements.jsonl (audit trail)
router/seller.py FastAPI agent that sells work — defines the requirements + work, wires one Paywall.gate()
router/buyer.py agent that auto-pays 402s (pay_and_get) with a client-side spend limit
router/eip3009.py the one crypto-subtle file: EIP-3009 typed-data, shared by buyer+facilitator so they can't drift
router/facilitator_mock.py in-process verify+settle (real sig check, mock chain) — the dry-run rail
router/facilitator_real.py live x402 facilitator over HTTP via the x402 SDK's own client; get_facilitator() picks mock↔real by config
router/config.py network/asset facts + .env loader; mock↔testnet↔mainnet is a one-line env change
run_demo.py end-to-end smoke (ephemeral keys, mock facilitator)
run_live.py live runner — settles a real x402 payment on Base Sepolia

Go live (Base Sepolia testnet)

  1. Two testnet wallets (buyer signs, seller receives):

    python3 -c "from eth_account import Account; a=Account.create(); print('addr', a.address); print('key', a.key.hex())"
    

    Do this twice. Keep the keys out of git (use .env).

  2. Fund the buyer with Base Sepolia testnet USDC (Circle faucet) + a little testnet ETH for any gas the facilitator relays.

  3. .env (copy .env.example.env, it's gitignored + auto-loaded):

    APV0_NETWORK=base-sepolia
    APV0_BUYER_PRIVATE_KEY=0x...        # buyer testnet key (signs)
    APV0_SELLER_ADDRESS=0x...           # seller address (receives)
    APV0_FACILITATOR_URL=https://x402.org/facilitator
    
  4. Run it live:

    python3 run_live.py
    

    The seller now uses RealFacilitator (the x402 SDK's own facilitator client), the buyer auto-pays the 402, and x402.org broadcasts the EIP-3009 USDC transfer on-chain. You get a real tx hash → https://sepolia.basescan.org/tx/<hash>.

    Gasless: with EIP-3009 the facilitator relays gas, so the buyer only needs testnet USDC, not ETH.

Gate your own endpoint (the product primitive)

from router.paywall import Paywall
from router.facilitator_real import get_facilitator

paywall = Paywall(get_facilitator(), my_requirements_fn)   # mock ↔ real by env

@app.get("/my-endpoint")
def my_endpoint(request: Request):
    return paywall.gate(request, lambda: do_expensive_work())

That one .gate() call handles the whole x402 V2 dance — 402 challenge → verify → settle → record receipt → deliver — and appends every settlement to the receipt-ledger. The deployer never touches the protocol or holds a key.

v1 dogfood — pay an agent, get REAL work back

v0 proved the payment. v1 proves the loop: a buyer agent pays → a seller agent does real inference → returns the result + a settlement receipt. The seller's _do_the_work(task) is the product's integration point (plug in your own model/compute/service).

APV0_OPENROUTER_KEY=sk-or-... python3 run_v1_dogfood.py "your task here"

V1 DOGFOOD PASS — agent paid agent for real work ✅ (mock payment + real model call by default; set APV0_NETWORK=base-sepolia for a real on-chain settle under the same loop). Without a key it returns a stub so the product still runs.

This is the authentic loop: an agent doing real work, metered + paid-for over x402 — the same primitive whether the work is a model call, a compute job, or any other billable agent task.

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured