Smart Router

Cheapest / fastest / safest path A to B across fiat on-ramps, bridges, and swap aggregators. No Claude tax. Pure heuristics. One call.

Overview

The router hits LI.FI, Socket/Bungee, and Across in parallel (8s timeout each), normalizes their responses into a unified step/fee shape, scores them with a mode-dependent formula, and returns a ranked list. Ramp quotes hit MoonPay / Ramp / Transak / Simplex / Mercuryo / Coinbase Pay / Binance Connect (stubbed pending partner keys). Results cache for 30s per-param.

All quotes are fed into an in-memory learnings ring buffer (cap 1000) that drives historical averages on /v1/route/pairs and live stats on /v1/route/learnings.

Endpoints

POST/v1/route/quote — best cross-chain route

POST/v1/route/ramp/quote — fiat on-ramp comparison

POST/v1/route/offramp/quote — fiat off-ramp comparison

POST/v1/route/execute — submit a route

GET/v1/route/status/:executionId — transitions submitted → confirming → confirmed

GET/v1/route/pairs — top 30 pre-computed pairs

GET/v1/route/learnings — ring buffer + aggregate stats

POST/v1/route/pcp-flow — headline: USD → PCP anywhere

POST /v1/route/quote

curl -X POST https://metercall.ai/v1/route/quote \
  -H 'content-type: application/json' \
  -d '{
    "from_chain": "base",
    "from_token": "USDC",
    "from_amount": 100,
    "to_chain": "arbitrum",
    "to_token": "USDC",
    "to_address": "0xYourWallet",
    "optimize": "cheapest"
  }'

Response:

{
  "routes": [
    {
      "id": "r_ln1_0",
      "steps": [
        { "type":"bridge","protocol":"across","from":"base","to":"arbitrum",
          "amount_in":100,"amount_out":99.78,"fee_usd":0.22,"ms_est":30000 }
      ],
      "total_fee_usd": 0.22,
      "total_ms_est": 30000,
      "net_output_amount": 99.78,
      "confidence": 0.95,
      "path_label": "via Across",
      "source": "live"
    }
  ],
  "recommended_id": "r_ln1_0",
  "explain": "chose via Across — 60% cheaper than via LI.FI",
  "optimize": "cheapest"
}

POST /v1/route/ramp/quote

curl -X POST https://metercall.ai/v1/route/ramp/quote \
  -H 'content-type: application/json' \
  -d '{
    "fiat_currency": "USD",
    "fiat_amount": 100,
    "target_chain": "base",
    "target_token": "USDC",
    "payment_method": "card"
  }'

POST /v1/route/pcp-flow

curl -X POST https://metercall.ai/v1/route/pcp-flow \
  -H 'content-type: application/json' \
  -d '{ "usd_amount": 100, "destination_chain": "arbitrum",
        "destination_wallet": "0xYourWallet" }'

Chains USD → cheapest on-ramp → USDC on Base → Uniswap V3 → PCP → bridge to destination_chain.

POST /v1/route/execute

{ "route_id": "r_ln1_0", "destination_wallet": "0xYourWallet" }
→ { "ok": true, "execution_id": "exec_...", "tx_hashes": [...], "status": "submitted" }

GET /v1/route/status/:executionId

Transitions with realistic timing: submitted (0–5s) → submitted (5–20s) → confirming (20–45s) → confirmed (> 45s).

Optimize modes

Mode	Formula	When to use
`cheapest`	`-fee_usd - (ms_est/60000) * 0.1`	Default. Ignores speed almost entirely.
`fastest`	`-ms_est - fee_usd * 0.1`	User is waiting. Fee is secondary.
`safest`	`protocol_safety * 100 - fee_usd * 0.5`	Large transfers. Audit status and liveness matter most.

Protocol safety weights: across=0.95, lifi_meta=0.92, stargate=0.9, socket_meta=0.88, hop=0.85, synapse=0.8, wormhole=0.75.

Response shape

Every route has steps[] with {type: "swap"|"bridge"|"ramp", protocol, from, to, amount_in, amount_out, fee_usd, ms_est}. Top-level carries total_fee_usd, total_ms_est, net_output_amount, confidence (0–1), path_label, and source (live | stub).

Execution flow

Client calls /v1/route/quote → selects a route_id.
Client calls /v1/route/execute with that route_id + destination_wallet.
Server returns an execution_id immediately.
Client polls /v1/route/status/:executionId every ~5s until status = confirmed.

Observability

Every quote writes to an in-memory ring buffer (cap 1000). /v1/route/learnings returns:

{
  "count": 832,
  "today_count": 214,
  "avg_fee_usd": 1.42,
  "fastest_ms": 28400,
  "total_value_routed_usd": 1203.88,
  "recent": [ { "pair":"base:USDC->arbitrum:USDC","chosen_protocol":"across","fee_usd":0.22,"ms":30000,"ts":... } ]
}

Every server-side quote is also logged as [ROUTE] from→to protocol fee_usd ms and every execution as [ROUTE_EXEC] id.

How we stay cheapest

Continuous learning loop via /v1/route/learnings — recent winners get visibility in /v1/route/pairs.
Five server-side L4 bots (chain health, gas, prices, TVL, protocol discovery) pre-populate global._l4Prices, global._l4GasLatest, global._l4ChainHealth, global._l4TopProtocols — no cold starts.
30-second per-param cache — burstable without hammering aggregators.
Parallel fan-out with 8s timeout + fail-soft stubs — UI never looks dead.

SDKs

<script src="https://metercall.ai/sdks/route.js"></script>
<script>
  const q = await MeterCallRoute.quote({
    from_chain: 'base', from_token: 'USDC', from_amount: 100,
    to_chain: 'arbitrum', to_token: 'USDC', optimize: 'cheapest'
  });
  console.log(q.recommended_id, q.explain);
</script>

Also works in Node 18+ via import('./sdks/route.js') (CommonJS / UMD wrapper).