Prototype · Preview: every server in the official registry, synced every 30 minutes. Reviews and accounts come in a later step.
PopNx MCPMCP servers for people who use AI · explained in plain words

AVEL

by Ensi81version 0.1.2i★ –iRuns on your computer
Registry nameio.github.Ensi81/avel

Install

Overview

Description from the maker (original English)

Identity/reputation verification for agent-to-agent payments across x402, AP2, MPP, checkout.

PopNx MCP’s three pillars

These three are never merged into one score: opinions, measurements and popularity are different things.

What you can do with it

We have not written a plain-language summary for this one yet. Below are the facts we read from the registry.

  • Connection: Runs on your computer
  • Transport: stdio
  • Category: Online shops & payments
  • You need to fill in: AVEL_PRIVATE_KEY

Install

Pick the AI app you use, then copy the settings.

Before you install

  • Basic checks: 1 of 1 passed
  • Runs on your computer with your user’s permissions (files, network, apps).
  • You fill in yourself: AVEL_BASE_URL, AVEL_PRIVATE_KEY

Values marked secret are keys to your account. Never share them or paste them into a chat.

You need

  • PyPIPython’s uv installed (for the uvx command)

Which app do you use?

Pick your app above. We remember it, so next time the button says “Install in …”.

Open Claude Desktop → Settings → Developer → Edit Config, and paste this into claude_desktop_config.json

{
  "mcpServers": {
    "avel": {
      "command": "uvx",
      "args": [
        "agent-interchange"
      ],
      "env": {
        "AVEL_BASE_URL": "your-value",
        "AVEL_PRIVATE_KEY": "your-value"
      }
    }
  }
}

Run this in your terminal

claude mcp add avel --env AVEL_BASE_URL=your-value --env AVEL_PRIVATE_KEY=your-value -- uvx agent-interchange

Fill in the empty values before running it.

Open in Cursor

Or copy the settings yourself

Paste into .cursor/mcp.json in your project, or ~/.cursor/mcp.json

{
  "mcpServers": {
    "avel": {
      "command": "uvx",
      "args": [
        "agent-interchange"
      ],
      "env": {
        "AVEL_BASE_URL": "your-value",
        "AVEL_PRIVATE_KEY": "your-value"
      }
    }
  }
}

Your browser asks to open the app, then the app shows the settings and asks you to confirm. Key values are left empty for you to fill in.

Open in VS Code

Or copy the settings yourself

Paste into .vscode/mcp.json in your project

{
  "servers": {
    "avel": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "agent-interchange"
      ],
      "env": {
        "AVEL_BASE_URL": "your-value",
        "AVEL_PRIVATE_KEY": "your-value"
      }
    }
  }
}

Your browser asks to open the app, then the app shows the settings and asks you to confirm. Key values are left empty for you to fill in.

This one runs on your computer, and ChatGPT can only connect to servers on the internet. Pick another app.

Your choice stays on this device in a cookie; no account needed.

Values you fill in

  • AVEL_BASE_URL env
    Base URL of the AVEL instance to verify against (default: https://aisrail.fly.dev)
  • AVEL_PRIVATE_KEY secret required env
    EVM private key of the agent identity this MCP server represents — signs every verify()/issue_challenge()/verify_challenge()/get_verdict() call. Never exposed as a tool argument.

These settings are generated from the maker’s registry entry. If the README says otherwise, follow the README.

What it can do

Runs on your computer; the tool list appears in your AI app after you install it.

Basic checksi

We check information anyone can see; we have not read the code. Checks that do not apply are left out of the count. · Checked Nov 7, 2026

  • ✓ Pass

    Maker matches the repo

    The registry name was verified with the GitHub account Ensi81, which owns the repo

  • … Not checked yet

    Still maintained

    We have not read this repo from GitHub yet

  • … Not checked yet

    Has a licence

    We have not read this repo from GitHub yet

  • – Not applicable

    Endpoint responds

    This one runs on your computer; there is no internet endpoint

  • – Not applicable

    Sign-in with OAuth

    Runs on your computer and uses local settings instead

  • … Not checked yet

    No hidden instructions in tool descriptions

    Runs on your computer; we can only read tool lists from internet endpoints

Compared with others in Online shops & paymentsi

  • More stars than 0% of the other 1753 servers in this category
  • Published to the registry more recently than 40% of the other 1753 servers in this category
  • Basic checks equal to or better than 100% of the other 1753 servers in this category

Compared with every active server in this category in the registry. Stars count only when the publisher owns the repo.

Common questions

Is AVEL safe?

It passed 1 of 1 applicable basic checks. We have not read the code. Before adding a key or token, see what access the server asks for and use a key with only the permissions you need.

What keys or codes does it need?

You need to fill in: AVEL_PRIVATE_KEY

Does it run on my computer or in the cloud?

It runs on your computer. Your AI app starts it when needed, and data does not pass through the maker’s servers (unless it calls an online service itself).

Is it still maintained?

The GitHub repo last changed –, and the latest version (0.1.2) reached the registry 2 months ago.

Which apps can use it?

Apps that support MCP, such as Claude Desktop, Claude Code, Cursor and VS Code. You also need Python (uv) on your computer.

What can I use instead?

In the same category: Spree Commerce: open-source B2B and marketplace ecommerce, Stripe MCP, MCP Cn Commerce

Alternatives

Servers in the same category from other makers, sorted by stars.

Reading a card:★ GitHub starsiThree pillars: reviews · checks · usageiUpdated = latest version published to the registryi

Stripe MCP

by stripe

Stripe’s official server lets AI view and manage customers, products, payments and invoices. Connect to Stripe’s cloud and sign in with OAuth.

Runs in the cloud★ 2,043MITUpdated 12 months ago
Install

MCP Cn Commerce

by TonyWang-hub

8 read-only MCP platform entries; support is operation-specific; merchant live acceptance pending

Runs on your computer★ 96MITUpdated 28 days ago
Install

README from the maker

Open the README

Copied from the maker’s repo as written. PopNx MCP has not reviewed or endorsed it. HTML, images and scripts are removed. Links lead off this site.

agent-interchange

SDK Python per AVEL (Agent Verification Layer) — verifica identità/reputazione di chi paga e chi riceve, su qualunque metodo di pagamento, senza mai custodire denaro. Il server è closed-source; questo pacchetto è lo strato client, pensato per essere pubblico e riusabile (licenza MIT). Standalone — nessuna dipendenza dal codice del server, solo httpx, eth-account e l'URL di un'istanza raggiungibile (https://aisrail.fly.dev).

Install

pip install -e .

Uso

from agent_interchange import InterchangeClient

client = InterchangeClient("https://aisrail.fly.dev") result = client.verify( payer_address=agent_a_address, payee_id="merchant-42", amount=25.0, currency="EUR", private_key=agent_a_private_key, fee_payment_method_id="pm_...", # con cosa si paga LA VERIFICA )

Il pagamento fra i due agenti non passa da qui. Questa chiamata dice soltanto se è sicuro procedere; poi l'agente paga sul rail che già usa. Noi incassiamo solo la provvigione per il controllo eseguito (1% dell'importo verificato, minimo 0,50 su carta) — dovuta anche su un verdetto DENIED, se la firma di chi chiama è comunque valida: è il prezzo del controllo, non dell'esito.

Tre esiti possibili:

RispostaSignificato
verdict: "APPROVED"Controllo superato, provvigione incassata: si procede
verdict: "DENIED"Non procedere. Provvigione comunque dovuta se la firma di chi chiama è valida — gratuito solo quando non lo è (nessuno da addebitare con certezza)
verdict_withheld: trueVerifica eseguita ma provvigione non ancora incassata: il verdetto (APPROVED o DENIED) resta trattenuto finché non si paga

private_key è la chiave EVM dell'agente (mai la nostra). Il primo uso autenticato lega quell'identità all'indirizzo che firma (trust-on-first-use): ogni chiamata successiva deve firmare con la STESSA chiave. Nessuna registrazione preventiva.

verify() usa protocol="native". Se l'agente sta già pagando con x402, AP2 o MPP non serve una seconda firma: si manda la prova che ha già prodotto per il pagamento — un esempio per ciascun protocollo è in scripts/protocol_adapters_demo.py.

Verifica simmetrica di B

Se il venditore (B) dichiara un indirizzo (payee_address), deve anche provare di controllarlo — altrimenti il verdetto è negato. La firma va calcolata da B (o da chi negozia per suo conto) su build_counterparty_message(payer_address, payee_id, amount, currency, nonce) e passata come payee_signature:

from agent_interchange.client import build_counterparty_message

message = build_counterparty_message(payer_address, payee_id, amount, currency, nonce) payee_signature = "0x" + Account.sign_message(encode_defunct(text=message), private_key=b_private_key).signature.hex()

result = client.verify( payer_address=a_address, payee_id="merchant-42", amount=25.0, currency="EUR", private_key=a_private_key, nonce=nonce, payee_address=b_address, payee_signature=payee_signature, )

Modello B-driven (protocollo challenge)

Più vicino a come funzionano davvero x402/MPP: è il venditore (B) a fissare prezzo e termini e a firmarli per primo, l'acquirente (A) li accetta firmando lo stesso testo.

# Lato B — nessuna chiamata di rete, va consegnato ad A fuori banda

challenge = client.issue_challenge( payee_address=b_address, payee_id="merchant-42", amount=25.0, currency="EUR", private_key=b_private_key, ttl_seconds=600, )

Lato A — accetta il challenge e chiede il verdetto in un colpo solo

result = client.verify_challenge( challenge, payer_address=a_address, private_key=a_private_key, fee_payment_method_id="pm_...", # o fee_method="x402" per pagare in stablecoin )

Il challenge è valido una sola volta per la transazione esatta che lo accetta: un ritentativo identico (necessario per completare il pagamento della commissione dopo un 402) resta ammesso, un riuso con termini diversi no.

Verdetto bidirezionale

Entrambe le parti possono leggere l'esito di una transazione specifica — non solo chi ha chiamato verify()/verify_challenge():

verdetto = client.get_verdict(

payee_address=b_address, nonce=nonce, address=b_address, private_key=b_private_key, # o quella di A )

Richiede una firma che provi di essere il pagante o il ricevente di quella transazione — a differenza dello storico aggregato (GET /ais/reputation/{address}, pubblico), il dettaglio di una singola transazione non lo è.

Caso marketplace

Quando payee_id/payee_address sono una vetrina e non chi fornisce davvero il prodotto/servizio, verify() accetta anche sub_merchant_address/sub_merchant_signature (stesso principio di B: la firma va calcolata su build_sub_merchant_message(payer_address, payee_id, amount, currency, nonce, sub_merchant_address)). Se dichiarato, la sua reputazione (sub_merchant_reputation nella risposta) entra nel verdetto tanto quanto quella di B.

Vedi example.py per una versione eseguibile del flusso base.

Metodi degli Agent Vaults — SPENTI

pay(), get_balance(), deposit() e withdraw() appartengono al modello precedente, in cui il servizio custodiva un saldo interno. Il server risponde ora 410 a quegli endpoint: non custodiamo più denaro, nemmeno per un istante. Il codice resta ed è riattivabile lato server con AGENT_VAULTS_ENABLED=true; la documentazione qui sotto è conservata come riferimento storico.

InterchangeClient

  • `verify(payer_address, payee_id, amount, currency, private_key, nonce="", payee_address=None, payee_signature=None, sub_merchant_address=None, sub_merchant_signature=None, fee_payment_method_id=None)` → dict. Il metodo da usare per il modello "A dichiara, B conferma". Firma EIP-191 sul testo canonico della transazione, chiede il verdetto e paga la provvigione. Un 402 non è un errore ma una risposta prevista ("verifica passata, ora paga"): viene restituito, non sollevato.
  • `issue_challenge(payee_address, payee_id, amount, currency, private_key, ttl_seconds=600)` → dict. Lato B (venditore), modello B-driven: emette e firma un challenge vincolante, nessuna chiamata di rete. Il risultato va consegnato ad A fuori banda.
  • `verify_challenge(challenge, payer_address, private_key, fee_payment_method_id=None, fee_method=None)` → dict. Lato A (acquirente): accetta il challenge di issue_challenge() e chiede il verdetto in un'unica chiamata.
  • `get_verdict(payee_address, nonce, address, private_key)` → dict. Verdetto bidirezionale: legge l'esito di una transazione specifica, utilizzabile da entrambe le parti coinvolte. Solleva InterchangeAPIError (403) se address non è né il pagante né il ricevente di quella transazione, (404) se non esiste alcun verdetto registrato per quel payee_address/nonce.
  • `get_agent_reputation` non è un metodo del client — è pubblico e senza autenticazione: GET {base_url}/ais/reputation/{address} con una semplice richiesta HTTP.
  • ~~`pay(payer_ref, payee_ref, amount, private_key, currency="USDC", task_id=None)`~~ *(spento, 410)* → PaymentResult. private_key signs an EIP-191 message binding this exact transaction (trust-on-first-use — see Usage above); mandatory since the Interchange added wallet authentication to /pay. Auto-derives agent_request_id from task_id + the call's own parameters (see client._derive_request_id) — pass your agent framework's own task/run id so a genuine retry (same task_id, same arguments) is safely idempotent, while a different payment under the same task_id never false-collides with it. Omitting task_id still works but loses the retry-safety property (a fresh random id is used, so a real retry would be treated as a brand new transaction).

Raises:

  • InsufficientBalanceError — payer's balance can't cover amount + fee.
  • RequestConflictError — HTTP 409: the derived id was already used for *different* content (a possible replay/hijack of a captured payment proof by untrusted middleware — rejected, not executed) or was reserved but never completed (never retried automatically).
  • InterchangeAPIError — any other non-2xx response, or the Interchange being unreachable. 401 means private_key doesn't match the signature; 403 means payer_ref is already bound to a different address (not this key).

A risk-score DENY (as opposed to an insufficient-balance one) is *not* raised — it comes back as a normal PaymentResult with .success == False; check .decision / .reasons for why.

  • `get_balance(agent_ref, private_key)` → dict[str, int], atomic units per currency, e.g. {"USDC": 500000}. private_key proves control of agent_ref (same trust-on-first-use binding as pay()); allowed with any key only if agent_ref was never used in pay()/ withdraw() yet.
  • `withdraw(agent_ref, amount_usdc, payout_address, private_key, task_id=None)` → dict. private_key is agent_ref's own key, proving it controls the balance being withdrawn — the Interchange's own treasury still signs and pays out on-chain, private_key here only authenticates the withdrawal request itself. Same task_id derivation and exception contract as pay() (InsufficientBalanceError on HTTP 402, RequestConflictError on 409, 401/403 same meaning as pay()'s); any other non-2xx (e.g. 403 for an unwhitelisted platform-treasury withdrawal) raises InterchangeAPIError.
  • `await deposit(agent_ref, amount_usdc, private_key, task_id=None)` → dict. Async (unlike the other three methods) and needs the deposit extra — pip install agent-interchange[deposit] — because it signs a real x402 payment with the depositing agent's own key. Calling it without the extra installed raises InterchangeAPIError (not an ImportError) with instructions. Base install (pip install -e .) never needs x402/eth-account unless you call this.

Verified live against a running Interchange (04 Sep 2026): a normal payment (fee correctly deducted from amount, not added on top); InsufficientBalanceError on an oversized pay()/withdraw(); an idempotent retry (same task_id + same arguments) returning the identical cached result with no second transfer/payout; the same task_id with *different* arguments executing as a distinct payment (no false conflict); a forged same-id-different-content request correctly rejected with RequestConflictError, simulating the replay/hijack scenario the Interchange's own P1 binding defends against (interchange/security.py in the main repo); a real on-chain withdraw() and deposit(), each with the resulting transaction confirmed on Base Sepolia; and deposit()'s clean-failure message when the deposit extra isn't installed (simulated by blocking the imports).

InterchangeTool — LangChain / CrewAI / AutoGen

from agent_interchange import InterchangeClient, InterchangeTool

client = InterchangeClient("http://localhost:8000") tool = InterchangeTool(client)

langchain_tools = tool.for_langchain() # requires: pip install langchain-core crewai_tools = tool.for_crewai() # requires: pip install crewai autogen_tools = tool.for_autogen() # requires: pip install autogen-core

Only wraps pay/get_balance as agent-callable tools — deliberately *not* deposit/withdraw: those move real funds and deposit needs a raw private key, both a heavier judgment call than exposing to an LLM's tool-calling loop by default. Call them directly on client from your own code instead.

Lazy imports: constructing InterchangeTool and calling pay()/ get_balance() on it directly never requires any of the three frameworks — only for_langchain()/for_crewai()/for_autogen() do, and each fails cleanly (ModuleNotFoundError) if its package isn't ins

Only the beginning is shown; read the rest in the repo

Ask PopNx