PacketExchange MCP server
io.packetexchange/mcppacketexchange.iostreamable-httpv2.1.0Scanned
SMS, voice calls, OTP verification, phone numbers and wholesale voice routes by Minutes Network.
33 tools: 6 can delete or cancel things, 8 do something that can't be undone, 6 change things and 13 only read.
- Reads private data · 0
- Reads text others write · 6
- Can send data out · 6
- 5 findings
Tools listed without sign-in. Remote posture
With Anlyon
6 tools here can delete or cancel things and 8 do something that can't be undone. Govern them.
- Anlyon lists the tools and adopts none.
- Each tool you adopt gets approvals and receipts.
- Your agent never holds the key.
A tool your agent calls directly is not routed through Anlyon and is not governed by it. Anlyon Radar only lists and labels, and these labels are the same either way.
What the PacketExchange MCP server can do
The PacketExchange MCP server is published by packetexchange.io, a domain the MCP Registry verified. It has 33 tools. It can delete or cancel things (6 tools: wmmn_topup_x402, switch_set_route_plan_entries, switch_assign_customer_routing, switch_bulk_upsert_supplier_rates and 2 more), do things that cannot be undone, such as sending, paying or publishing (wmmn_purchase_route, wmmn_send_sms, wmmn_make_call, wmmn_send_voice_otp and 4 more) and create or change data (6 tools: wmmn_verify_check, switch_create_customer, switch_create_supplier, switch_ar_aging and 2 more). 13 tools only read.
6 tools bring in text other people can write (wmmn_send_sms, wmmn_make_call, wmmn_get_call and 3 more); 6 tools can send data out of your system (wmmn_send_sms, wmmn_make_call, wmmn_send_voice_otp and 3 more).
The fixed checks matched 5 findings (long-description, credentials-mention and annotation-conflict). Each one comes with the text it matched.
Its remote endpoint lists its tools without sign-in, and its tools changed once since Anlyon Radar first saw them.
Last scanned Nov 17, 2026, version 2.1.0.
Tools
Grouped by what each tool does, most consequential first. Open one for its reasons and full description; descriptions are the publisher's own words.
33 tools
Deletes or cancels 6Claude Code: deny
switch_assign_customer_routingMUTATES STATE.No findings
Why it is in this group
- The name "switch_assign_customer_routing" says "assign": a word for a change.
- The server marks it destructive (destructiveHint: true).
Description, as the server wrote it MUTATES STATE. Point a customer at its routing: dialplan, route group, failover route group, LCR strategy override (cheapest|best_quality|balanced), or a direct supplier (one marketplace route OR one vendor trunk, not both). Only the fields you pass change: omitted fields keep their current value, and null clears a field. Writes the customer's first trunk (a customer with no trunk is refused); multi-trunk customers are configured per trunk in the dashboard. Requires an active Switch subscription (402 SWITCH_SUBSCRIPTION_REQUIRED otherwise) and a full-access API key (scoped keys need switch:manage). Test keys are NOT sandboxed here: they read and change your live Switch data.
Annotations the server declares
- title: Switch: set a customer's routing
- readOnlyHint: false
- destructiveHint: true
- idempotentHint: true
- openWorldHint: false
switch_bulk_upsert_sell_ratesMUTATES STATE.No findings
Why it is in this group
- The name "switch_bulk_upsert_sell_rates" says "upsert": a word for a change.
- The server marks it destructive (destructiveHint: true).
Description, as the server wrote it MUTATES STATE. Bulk upsert the operator's retail SELL deck (A-Z), de-duped last-wins, up to 50000 rows. Fires the sell_rate.changed webhook. The efficient way to load or refresh retail pricing. Requires an active Switch subscription (402 SWITCH_SUBSCRIPTION_REQUIRED otherwise) and a full-access API key (scoped keys need switch:manage). Test keys are NOT sandboxed here: they read and change your live Switch data.
Annotations the server declares
- title: Switch: load sell rates
- readOnlyHint: false
- destructiveHint: true
- idempotentHint: true
- openWorldHint: false
switch_bulk_upsert_supplier_ratesMUTATES STATE.No findings
Why it is in this group
- The name "switch_bulk_upsert_supplier_rates" says "upsert": a word for a change.
- The server marks it destructive (destructiveHint: true).
Description, as the server wrote it MUTATES STATE. Bulk upsert a supplier trunk's A-Z COST deck (paste/CSV rows), de-duped last-wins, up to 50000 rows. Price increases may be queued until the supplier's rate-notice period ends instead of applying at once. The efficient way to load or refresh vendor pricing. Requires an active Switch subscription (402 SWITCH_SUBSCRIPTION_REQUIRED otherwise) and a full-access API key (scoped keys need switch:manage). Test keys are NOT sandboxed here: they read and change your live Switch data.
Annotations the server declares
- title: Switch: load supplier rates
- readOnlyHint: false
- destructiveHint: true
- idempotentHint: true
- openWorldHint: false
switch_set_route_plan_entriesMUTATES STATE.No findings
Why it is in this group
- The name "switch_set_route_plan_entries" says "set": a word for a change.
- The server marks it destructive (destructiveHint: true).
Description, as the server wrote it MUTATES STATE. Replace ALL entries of a route group with an ordered supplier list (per-prefix, with failover priority, load-balance weight, and ASR/PDD gates). The primary tool to (re)configure LCR/failover routing. Enforces exactly-one-target (routeId XOR vendorTrunkId) and ownership. Requires an active Switch subscription (402 SWITCH_SUBSCRIPTION_REQUIRED otherwise) and a full-access API key (scoped keys need switch:manage). Test keys are NOT sandboxed here: they read and change your live Switch data.
Annotations the server declares
- title: Switch: replace a route plan
- readOnlyHint: false
- destructiveHint: true
- idempotentHint: true
- openWorldHint: false
switch_update_fraud_settingsMUTATES STATE.No findings
Why it is in this group
- The name "switch_update_fraud_settings" says "update": a word for a change.
- The server marks it destructive (destructiveHint: true).
Description, as the server wrote it MUTATES STATE. Update per-operator fraud screening: enable/disable, block/alert score thresholds (1-100), and high-risk destination prefixes. highRiskPrefixes REPLACES the whole list; an empty list reverts to the platform defaults. Tune protection. Requires an active Switch subscription (402 SWITCH_SUBSCRIPTION_REQUIRED otherwise) and a full-access API key (scoped keys need switch:manage). Test keys are NOT sandboxed here: they read and change your live Switch data.
Annotations the server declares
- title: Switch: update fraud screening
- readOnlyHint: false
- destructiveHint: true
- idempotentHint: true
- openWorldHint: false
wmmn_topup_x402SPENDS MONEY FROM AN ON-CHAIN WALLET.Reads text others writeNo findings
Why it is in this group
- The name "wmmn_topup_x402" says "topup": a word that usually means money moves, which cannot be taken back.
- The server marks it destructive (destructiveHint: true).
What it can reach
- untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
Description, as the server wrote it SPENDS MONEY FROM AN ON-CHAIN WALLET. Funds your PacketExchange balance with real USDC on Base over the x402 protocol. Step 1 is safe/read-only: call with just amount_usd to receive the payment requirements (HTTP 402 challenge naming the network, USDC amount and payTo address; nothing is charged). Step 2 MOVES FUNDS: sign the EIP-3009 USDC transferWithAuthorization, base64-encode it, and call again passing it as payment_header; the transfer settles on-chain and your balance is credited. Amounts are USD, min $5, max $50,000, subject to account verification (KYC) and top-up velocity limits. The credit goes to your LIVE balance whichever key you use: a test key does not simulate this. Requires the billing:write scope.
Annotations the server declares
- title: Top up with USDC (x402)
- readOnlyHint: false
- destructiveHint: true
- idempotentHint: false
- openWorldHint: true
Can't be undone 8Claude Code: ask
switch_analytics_breakdownTop-100 metric breakdown grouped by customer, supplier, or destination (the first 3 digits of the dialled number).Can send data out1 finding
Why it is in this group
- The input addresses people (input.to): what it sends cannot be taken back.
- The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
What it can reach
- outboundThe input takes a destination (input.to): what the tool writes goes there.
Findings
- Warn
The server marks "switch_analytics_breakdown" read-only (readOnlyHint: true), and its own name, description or input give the class "irreversible".
What this check looks forMatched text
annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":false}; The input addresses people (input.to): what it sends cannot be taken back.
Description, as the server wrote it Top-100 metric breakdown grouped by customer, supplier, or destination (the first 3 digits of the dialled number). Find the most/least profitable customers, worst-quality suppliers, or costliest destinations. Requires an active Switch subscription (402 SWITCH_SUBSCRIPTION_REQUIRED otherwise) and a full-access API key (scoped keys need switch:manage). Test keys are NOT sandboxed here: they read and change your live Switch data.
Annotations the server declares
- title: Switch: breakdown by customer, supplier or destination
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
switch_analytics_overviewUnified P&L + quality KPIs (revenue, cost, margin, marginPct, calls, answered, minutes, ASR, ACD, SMS count, avg PDD) across all operator CDR traffic for a date range.Can send data out1 finding
Why it is in this group
- The input addresses people (input.to): what it sends cannot be taken back.
- The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
What it can reach
- outboundThe input takes a destination (input.to): what the tool writes goes there.
Findings
- Warn
The server marks "switch_analytics_overview" read-only (readOnlyHint: true), and its own name, description or input give the class "irreversible".
What this check looks forMatched text
annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":false}; The input addresses people (input.to): what it sends cannot be taken back.
Description, as the server wrote it Unified P&L + quality KPIs (revenue, cost, margin, marginPct, calls, answered, minutes, ASR, ACD, SMS count, avg PDD) across all operator CDR traffic for a date range. The go-to health snapshot. Requires an active Switch subscription (402 SWITCH_SUBSCRIPTION_REQUIRED otherwise) and a full-access API key (scoped keys need switch:manage). Test keys are NOT sandboxed here: they read and change your live Switch data.
Annotations the server declares
- title: Switch: P&L and quality overview
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
switch_create_invoiceTwo steps, like a human review.No findings
Why it is in this group
- The name "switch_create_invoice" says "create": a word for a change.
- The description says "irreversible".
Description, as the server wrote it Two steps, like a human review. Step 1 (read-only, writes nothing): call WITHOUT review_token to get a preview of the AR invoice for a customer over a period from settled CDR usage (itemized voice + SMS lines, totals, anything unrated) plus a reviewToken. Step 2 (MUTATES STATE, irreversible): call again with the SAME arguments and review_token to issue it; this allocates the invoice number, claims the usage and emits invoice.created. If the usage changed since the preview the issue is refused (409) with a fresh preview. Issuing does not send the invoice to the customer. Requires an active Switch subscription (402 SWITCH_SUBSCRIPTION_REQUIRED otherwise) and a full-access API key (scoped keys need switch:manage). Test keys are NOT sandboxed here: they read and change your live Switch data.
Annotations the server declares
- title: Switch: preview and issue an invoice
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: false
wmmn_make_callSPENDS MONEY.Reads text others writeCan send data outNo findings
Why it is in this group
- The name "wmmn_make_call" says "call": a word for a change.
- The input addresses people (input.to): what it sends cannot be taken back.
What it can reach
- untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
- outboundThe input takes a destination (input.to): what the tool writes goes there.
Description, as the server wrote it SPENDS MONEY. Places a REAL outbound phone call and CHARGES your balance for a connected call: billable minutes (per the route billing increment) at the route price, plus a platform fee of 2% of the carrier cost, capped at $0.001 per call or message (0.5%, same cap, when the traffic runs over your own SIP trunk; no fee when the carrier cost is 0). Unanswered calls cost nothing. The worst-case cost for max_duration is reserved from your balance up front and the unused part refunded on hangup. The call is placed ASYNCHRONOUSLY: this tool waits up to wait_seconds (default 20) for it to finish, then returns either the final result (status completed, no_answer, busy or failed, with duration, cost, plain-words hangup reason and any gathered digits) or, if the call is still running, its call_id and live status - check it later with wmmn_get_call. Optional actions make the answered call speak (say, in en/es/fr/de/pt/hi), play an https MP3, collect keypad digits (gather), pause or hang up, in order; the call ends when they finish; no charge for the speech. Omit route_id to Smart-Route across the public marketplace (no purchase needed); a route_id must be a route you have purchased. A test key (wmmn_test_sk_...) simulates against test credit with no real telephony and runs no actions. Requires the voice:send scope.
Annotations the server declares
- title: Place a voice call
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: true
wmmn_purchase_routePROVISIONING ACTION (not read-only).1 finding
Why it is in this group
- The name "wmmn_purchase_route" says "purchase": a word that usually means money moves, which cannot be taken back.
Findings
- Info
The description mentions "credentials".
What this check looks forMatched text
description: "...ront charge, but it provisions live SIP credentials, and any traffic you then send over the..."
Description, as the server wrote it PROVISIONING ACTION (not read-only). Connect a marketplace route to your account. Connecting is free, with no upfront charge, but it provisions live SIP credentials, and any traffic you then send over the route CHARGES your balance per minute/message at the route price. Requires a LIVE key with the purchases:write scope: a test key (wmmn_test_sk_...) is refused with 400 "Route purchases require a LIVE API key". Your email must be verified; a private route needs an access grant from its seller; US voice routes need a complete STIR/SHAKEN profile and start as pending_review; connecting a route you already hold returns 409. The response includes your SIP username and password, shown once here (also retrievable later with wmmn_list_purchases on a full-access key).
Annotations the server declares
- title: Connect a marketplace route
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: false
wmmn_send_smsSPENDS MONEY.Reads text others writeCan send data outNo findings
Why it is in this group
- The name "wmmn_send_sms" says "send": a word for sending, and what is sent cannot be taken back.
- The input addresses people (input.to): what it sends cannot be taken back.
What it can reach
- untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
- outboundThe name "wmmn_send_sms" says "send": it sends or publishes something to other people.
- outboundThe input takes a destination (input.to): what the tool writes goes there.
Description, as the server wrote it SPENDS MONEY. Sends a REAL SMS and CHARGES your balance: the route price per segment, plus a platform fee of 2% of the carrier cost, capped at $0.001 per call or message (0.5%, same cap, when the traffic runs over your own SIP trunk; no fee when the carrier cost is 0). Long messages are split into segments (160 GSM-7 characters, 153 per part when split; 70 Unicode, 67 per part). Returns message id, segment count and exact USD cost. The returned status is carrier ACCEPTANCE, not handset delivery; delivery receipts are not returned by this tool. Omit route_id to Smart-Route across the public marketplace (no purchase needed); a route_id must be a route you have purchased. A test key (wmmn_test_sk_...) simulates against test credit with no real delivery. Requires the sms:send scope.
Annotations the server declares
- title: Send an SMS
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: true
wmmn_send_voice_otpSPENDS MONEY.Reads text others writeCan send data outNo findings
Why it is in this group
- The name "wmmn_send_voice_otp" says "send": a word for sending, and what is sent cannot be taken back.
- The input addresses people (input.to): what it sends cannot be taken back.
- The input takes a statement or command to run (input.code), which can change anything it can reach.
What it can reach
- untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
- outboundThe name "wmmn_send_voice_otp" says "send": it sends or publishes something to other people.
- outboundThe input takes a destination (input.to): what the tool writes goes there.
Description, as the server wrote it SPENDS MONEY. Places a REAL short phone call (capped at 120 seconds) and speaks a one-time passcode digit by digit ("Your {brand} verification code is 4, 8, 2, 9, 1, 3. Again, ..."), repeated (default twice), then hangs up. Billed like any API call: billable minutes at the route price, plus a platform fee of 2% of the carrier cost, capped at $0.001 per call or message (0.5%, same cap, when the traffic runs over your own SIP trunk; no fee when the carrier cost is 0); no text-to-speech surcharge. Returns as soon as the call is dialled (status "initiated") with a voice_otp id; the final outcome and cost are recorded later (GET /api/v1/comms/voice-otp/{id} in the REST API). Omit code to have one generated; it is returned only when return_code is true. Limits: 5 per number per hour, 30 second resend cooldown, 20 per number range per hour, 1000 per account per day; embargoed, high-risk and do-not-contact numbers are refused. A test key (wmmn_test_sk_...) simulates with no real call and returns status "accepted" with simulated: true. Requires the voice:send scope. For send-and-check, prefer wmmn_verify_start with channel "voice".
Annotations the server declares
- title: Send a one-time code by voice call
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: true
wmmn_verify_startSPENDS MONEY.Reads text others writeCan send data outNo findings
Why it is in this group
- The name "wmmn_verify_start" says "start": a word for a change.
- The input addresses people (input.to): what it sends cannot be taken back.
What it can reach
- untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
- outboundThe input takes a destination (input.to): what the tool writes goes there.
Description, as the server wrote it SPENDS MONEY. Starts a phone-number verification: generates a one-time code, stores only a keyed hash of it, and sends it by SMS or by a voice call that speaks it. Billed exactly like a normal SMS or API call: the route price, plus a platform fee of 2% of the carrier cost, capped at $0.001 per call or message (0.5%, same cap, when the traffic runs over your own SIP trunk; no fee when the carrier cost is 0). Returns a verification_id (never the code on a live key). The code expires after expiry_seconds (default 600), allows 5 check attempts and approves once. Rate limited: 5 starts per number per hour, 30 second resend cooldown, 20 per number range per hour, and 1000 per account per day; embargoed, high-risk premium and do-not-contact destinations are refused. A test key simulates the send and returns testCode so the flow can be completed. Requires the verify:write scope.
Annotations the server declares
- title: Start a 2FA verification
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: true
Changes things 6Claude Code: ask
switch_ar_agingAR aging summary: total outstanding, open count, and current/1-30/31-60/61-90/90+ buckets across open/partial/overdue invoices.No findings
Why it is in this group
- Nothing in its name or description says what it does, so it is counted as a write.
- The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
Description, as the server wrote it AR aging summary: total outstanding, open count, and current/1-30/31-60/61-90/90+ buckets across open/partial/overdue invoices. Key collections-health read. Requires an active Switch subscription (402 SWITCH_SUBSCRIPTION_REQUIRED otherwise) and a full-access API key (scoped keys need switch:manage). Test keys are NOT sandboxed here: they read and change your live Switch data.
Annotations the server declares
- title: Switch: receivables aging
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
switch_create_customerMUTATES STATE.No findings
Why it is in this group
- The name "switch_create_customer" says "create": a word for a change.
Description, as the server wrote it MUTATES STATE. Create a switch-native customer and mint its first API key (returned ONCE, with sipPassword). Set markupPct, credit limit, margin floor, spend and concurrency caps, CPS, blocked prefixes, tax details and portalEmail. externalRef is an idempotency key: repeating a create with the same externalRef returns the existing customer WITHOUT a new API key or sipPassword. The provisioning entry point for onboarding a buyer. Requires an active Switch subscription (402 SWITCH_SUBSCRIPTION_REQUIRED otherwise) and a full-access API key (scoped keys need switch:manage). Test keys are NOT sandboxed here: they read and change your live Switch data.
Annotations the server declares
- title: Switch: create a customer
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: false
switch_create_payableMUTATES STATE.No findings
Why it is in this group
- The name "switch_create_payable" says "create": a word for a change.
Description, as the server wrote it MUTATES STATE. Generate an AP payable for a supplier trunk over a period from CDR seller-credit cost. The core settlement action for what the operator owes vendors. Requires an active Switch subscription (402 SWITCH_SUBSCRIPTION_REQUIRED otherwise) and a full-access API key (scoped keys need switch:manage). Test keys are NOT sandboxed here: they read and change your live Switch data.
Annotations the server declares
- title: Switch: create a supplier payable
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: false
switch_create_supplierMUTATES STATE.No findings
Why it is in this group
- The name "switch_create_supplier" says "create": a word for a change.
Description, as the server wrote it MUTATES STATE. Create an external supplier vendor trunk (SIP host/port/transport, auth, codecs, default rate, capacity, CPS, settlement mode) and/or its SMS delivery (HTTP URL or SMPP bind). Needs a label plus at least one way to deliver traffic: sipHost, smsDeliveryUrl, or smppHost with smppSystemId. A new trunk may be held as draft until it passes readiness checks. The provisioning entry point for a termination vendor. Requires an active Switch subscription (402 SWITCH_SUBSCRIPTION_REQUIRED otherwise) and a full-access API key (scoped keys need switch:manage). Test keys are NOT sandboxed here: they read and change your live Switch data.
Annotations the server declares
- title: Switch: add a supplier trunk
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: false
switch_run_nettingMUTATES STATE.No findings
Why it is in this group
- The name "switch_run_netting" says "run": a word for a change.
Description, as the server wrote it MUTATES STATE. Run bilateral netting for a counterparty over a period: offsets AR (their invoices) against AP (their payables), returning net amount and direction (settled|receivable|payable). Key settlement automation. Requires an active Switch subscription (402 SWITCH_SUBSCRIPTION_REQUIRED otherwise) and a full-access API key (scoped keys need switch:manage). Test keys are NOT sandboxed here: they read and change your live Switch data.
Annotations the server declares
- title: Switch: net off what you and a counterparty owe
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: false
wmmn_verify_checkFree, no send.No findings
Why it is in this group
- The input takes a statement or command to run (input.code), which can change anything it can reach.
Description, as the server wrote it Free, no send. Checks a code the user entered against a verification from wmmn_verify_start. Returns status approved | denied | expired | max_attempts and attemptsRemaining. Only "approved" proves the user holds the number. Each wrong code uses one of 5 attempts; an approved verification cannot be approved again (a repeat check is denied with reason already_used). Check with the same kind of key (test or live) that started the verification. Requires the verify:write scope.
Annotations the server declares
- title: Check a 2FA code
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: false
Only reads 13Show 13 read-only tools
switch_fraud_eventsList the 100 most recent fraud events (newest first) with score, action (flag|block), destination, reasons, and customer label.No findings
Why it is in this group
- The description's first verb is "list": a word for reading.
Description, as the server wrote it List the 100 most recent fraud events (newest first) with score, action (flag|block), destination, reasons, and customer label. Monitor and triage abuse. Requires an active Switch subscription (402 SWITCH_SUBSCRIPTION_REQUIRED otherwise) and a full-access API key (scoped keys need switch:manage). Test keys are NOT sandboxed here: they read and change your live Switch data.
Annotations the server declares
- title: Switch: recent fraud events
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
switch_list_customersList all sell-side customers (sub-accounts) with balance, credit limit, markup, margin floor, spend caps, concurrency, and status.No findings
Why it is in this group
- The name "switch_list_customers" says "list": a word for reading.
Description, as the server wrote it List all sell-side customers (sub-accounts) with balance, credit limit, markup, margin floor, spend caps, concurrency, and status. Core inventory read for who is connected. Requires an active Switch subscription (402 SWITCH_SUBSCRIPTION_REQUIRED otherwise) and a full-access API key (scoped keys need switch:manage). Test keys are NOT sandboxed here: they read and change your live Switch data.
Annotations the server declares
- title: Switch: list customers
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
switch_list_invoicesList AR invoices with customer/status filters, pagination, and computed overdue flag.No findings
Why it is in this group
- The name "switch_list_invoices" says "list": a word for reading.
Description, as the server wrote it List AR invoices with customer/status filters, pagination, and computed overdue flag. Survey receivables. Requires an active Switch subscription (402 SWITCH_SUBSCRIPTION_REQUIRED otherwise) and a full-access API key (scoped keys need switch:manage). Test keys are NOT sandboxed here: they read and change your live Switch data.
Annotations the server declares
- title: Switch: list invoices
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
switch_list_suppliersList all buy-side supplier vendor trunks with SIP egress config, default rate, capacity, CPS, settlement mode, and status.No findings
Why it is in this group
- The name "switch_list_suppliers" says "list": a word for reading.
Description, as the server wrote it List all buy-side supplier vendor trunks with SIP egress config, default rate, capacity, CPS, settlement mode, and status. Core inventory read for termination options. Requires an active Switch subscription (402 SWITCH_SUBSCRIPTION_REQUIRED otherwise) and a full-access API key (scoped keys need switch:manage). Test keys are NOT sandboxed here: they read and change your live Switch data.
Annotations the server declares
- title: Switch: list suppliers
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
switch_preview_routingRead-only routing preview for a dialed number: resolves the dialplan (matched rule, block verdict, translated number, out-CLI) and returns the ordered supplier failover chain with …No findings
Why it is in this group
- The name "switch_preview_routing" says "preview": a word for reading.
Description, as the server wrote it Read-only routing preview for a dialed number: resolves the dialplan (matched rule, block verdict, translated number, out-CLI) and returns the ordered supplier failover chain with per-attempt price and prefix match. No side effects: the single best tool to sanity-check routing before real traffic. Optionally scope to a customer via sub_account_id (their assigned dialplan) and customer_trunk_id (that trunk's restrictions; default is the customer's single active trunk). Requires an active Switch subscription (402 SWITCH_SUBSCRIPTION_REQUIRED otherwise) and a full-access API key (scoped keys need switch:manage). Test keys are NOT sandboxed here: they read and change your live Switch data.
Annotations the server declares
- title: Switch: preview routing for a number
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
wmmn_get_balanceRead-only.No findings
Why it is in this group
- The name "wmmn_get_balance" says "get": a word for reading.
Description, as the server wrote it Read-only. Current account balance and test credit, in USD, plus their sum for reference. Only the balance pays for live calls and messages; test credit covers test API keys (simulated traffic) only. Requires the account:read scope (full-access keys have it).
Annotations the server declares
- title: Account balance
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
wmmn_get_callReturns one call placed with wmmn_make_call (or POST /comms/calls): live status (queued, ringing, answered, then completed, no_answer, busy or failed), timestamps, duration, exact …Reads text others writeNo findings
Why it is in this group
- The name "wmmn_get_call" says "get": a word for reading.
What it can reach
- untrusted-inputThe description's first sentence says "return" and "post": it reads posts, which other people can write.
Description, as the server wrote it Returns one call placed with wmmn_make_call (or POST /comms/calls): live status (queued, ringing, answered, then completed, no_answer, busy or failed), timestamps, duration, exact USD cost, the hangup reason in plain words, and the keypad digits collected by gather actions. Free. Requires the voice:send scope.
Annotations the server declares
- title: Get call status
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
wmmn_list_purchasesRead-only.1 finding
Why it is in this group
- The name "wmmn_list_purchases" says "list": a word for reading.
Findings
- Info
The description mentions "credentials".
What this check looks forMatched text
description: "...atus, per-unit USD pricing and your SIP credentials (username + password) for each. Require..."
Description, as the server wrote it Read-only. List the routes you have connected (your purchases, newest 50), with status, per-unit USD pricing and your SIP credentials (username + password) for each. Requires a FULL-ACCESS API key (one created with no scope restrictions): because the response contains SIP passwords, every scoped key is refused with 403.
Annotations the server declares
- title: List connected routes
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
wmmn_list_routesRead-only, free, no key needed.1 finding
Why it is in this group
- The name "wmmn_list_routes" says "list": a word for reading.
Findings
- Info
The description is 2,273 characters long, over the 2,000 this check looks for.
What this check looks forMatched text
description ends: "...from origin_group; other calls go to your next route. Filter with origin_policy."
Description, as the server wrote it Read-only, free, no key needed. Search and filter live voice/SMS termination routes on the PacketExchange marketplace. By default it returns what the website shows: one row per route and mobile network, routes we have PROVEN first (tested_desc), each with two platform-measured marks: route_status (proven = a real call or our test connected to that network recently; weak = it connects but few calls do; at_risk = the latest test failed; untested = no evidence yet, not a fault) and caller_id (full = the caller ID arrived intact on a real phone on that network; replaced; none; untested). Every listing also carries a verification tier for the whole route (owner, 16 Oct 2026), and every view lists the best tier first: cli = "International CLI" (our international caller-ID test showed the exact number we sent, in full), ncli = "NCLI" (the test reached the phone with a different number or none), connected = "Qualified" (a real answered call or passed test, no caller-ID result), not_verified = "Untested" (not tested by us yet, or the latest test failed; new listings start here; still buyable). Quoted names are what the marketplace shows; the values are stable keys. Use verification=<tier> to keep one tier, or tested=true / verified_cli=true to keep only proven rows. Prices are USD per minute (voice) or per message (SMS). ASR/ACD figures are the SELLER-DECLARED expected values, not platform-measured. Route quality tiers (route_type): direct | premium | standard | ncli (NCLI = no caller ID: lowest cost, more intermediate carriers, and the called party does not see the number). Set per_network=false for one row per listing (a multi-destination A-Z deck then shows only its cheapest "from" rate, so check wmmn_route_details). CALLER ID ORIGIN: some sellers price by where the call originates, judged by the caller ID sent (From and P-Asserted-Identity). origin_policy "dual" = two prices: origin_price applies when the call presents a VALID caller ID from origin_group (EEA = the 27 EU countries, Iceland, Liechtenstein, Norway), price applies to every other call (another country, no caller ID, or an invalid one). "origin_only" = the route carries only calls with a caller ID from origin_group; other calls go to your next route. Filter with origin_policy.
Annotations the server declares
- title: Search marketplace routes
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
wmmn_list_transactionsRead-only.No findings
Why it is in this group
- The name "wmmn_list_transactions" says "list": a word for reading.
Description, as the server wrote it Read-only. Recent billing transactions for your account, newest first, amounts in USD (charges are negative, credits/top-ups positive). Types: topup, charge, credit, payout, platform_fee, test_credit. Requires the account:read scope.
Annotations the server declares
- title: Billing transactions
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
wmmn_market_summaryRead-only, free.No findings
Why it is in this group
- The description's first verb is "read": a word for reading.
Description, as the server wrote it Read-only, free. Live marketplace snapshot: total routes, destinations, voice and SMS route counts, market status and as-of time. When the connection carries an API key, the averages are included too (capacity, average and measured ASR, ACD and USD price); without a key only the counts are returned. 24h trading volumes are not part of this summary.
Annotations the server declares
- title: Marketplace summary
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
wmmn_resolve_routeRead-only, free, no call placed.No findings
Why it is in this group
- The name "wmmn_resolve_route" says "resolve": a word for reading.
Description, as the server wrote it Read-only, free, no call placed. Preview which route Smart Routing would pick for a destination, plus up to 4 ranked alternatives. It ranks every active public marketplace route for that destination together with any private routes you have purchased. Requires an API key with the routes:read scope (full-access keys have it).
Annotations the server declares
- title: Preview Smart Routing for a number
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
wmmn_route_detailsRead-only, free, no key needed.No findings
Why it is in this group
- The description's first verb is "read": a word for reading.
Description, as the server wrote it Read-only, free, no key needed. Full details for one route — USD pricing, billing increment, seller-declared expected ASR/ACD/PDD, CLI type, capacity, and the first-party Exchange Score plus seller trust profile. originPolicy / originGroup / originPricing say whether the route is priced by caller ID origin: with "dual", originPricing.originFrom is the lowest price for calls presenting a valid caller ID from originGroup (EEA), and pricePerUnit / the deck rates are the International price for every other call; "origin_only" carries only calls with a caller ID from originGroup.
Annotations the server declares
- title: Route details
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
This is not a safety rating, and no finding is not a safety guarantee. The groups and findings come from fixed checks of the tool definitions the server returned on the date shown, not from its code or what it does when called. How the checks work.
Exfil surface
Which tools read private data, which read text someone else can write, and which can send data out. One of each is a path a prompt injection can use.
No private-data, untrusted-input, outbound path in this server alone: no tool is tagged private-data.
- Reads private data0
- Reads files, mail, messages, databases, secrets, calendars or repositories.
- Reads text others write6
- Brings in text someone else can write: web pages, inbound email, messages, issues and comments.
- Can send data out6
- Can carry data to people or systems: it sends, posts or publishes, or takes a recipient, channel or URL.
Other servers in the same agent add their own tools to these paths. Audit your whole setup to see them together.
Remote posture
What anyone on the internet sees at the remote endpoint: an anonymous initialize and tools/list, and the public OAuth metadata.
Looked at Nov 20, 2026, 19:14 UTC, unchanged since Nov 20, 2026. No credentials were sent and no tool was called. What is looked at.
- Reachable
- yes
- HTTP status of an anonymous initialize
- 200
- tools/list served without credentials
- yes
- Tools that change things, listed without credentials
- 31
- Protocol version
- 2025-06-18
- WWW-Authenticate on 401
- not observed
- resource_metadata in WWW-Authenticate
- not observed
- RFC 9728 protected resource metadata
- https://packetexchange.io/.well-known/oauth-protected-resource/mcp/http
- Authorization servers
- https://packetexchange.io
- Legacy HTTP+SSE transport
- no
- Dynamic client registration
- yes
- Client ID metadata documents
- no
- HighThe whole server
The endpoint served tools/list without credentials, and the list has 31 tools that change things (7 destructive, 13 irreversible, 11 write). Whether a call also works without credentials was not tested: tools are never called.
What this check looks forMatched text
https://packetexchange.io/mcp/http: tools/list returned 48 tools to an anonymous client, including px_topup_x402 (destructive), switch_set_route_plan_entries (destructive), switch_assign_customer_routing (destructive), switch_bulk_upsert_supplier_rates (destructive), switch_bulk_upsert_sell_rates (destructive) and 26 more
Posture changes
- Nov 20, 2026
- Tools that change things, listed without credentials: 20 to 31
- RFC 9728 protected resource metadata: none to https://packetexchange.io/.well-known/oauth-protected-resource/mcp/http
- Authorization servers: none to https://packetexchange.io
- Dynamic client registration: not observed to yes
- Client ID metadata documents: not observed to no
Findings
Matches of fixed checks, with the text they matched.
The server marks "switch_analytics_overview" read-only (readOnlyHint: true), and its own name, description or input give the class "irreversible".
What this check looks forMatched text
annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":false}; The input addresses people (input.to): what it sends cannot be taken back.
The server marks "switch_analytics_breakdown" read-only (readOnlyHint: true), and its own name, description or input give the class "irreversible".
What this check looks forMatched text
annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":false}; The input addresses people (input.to): what it sends cannot be taken back.
- Infowmmn_list_routes
The description is 2,273 characters long, over the 2,000 this check looks for.
What this check looks forMatched text
description ends: "...from origin_group; other calls go to your next route. Filter with origin_policy."
The description mentions "credentials".
What this check looks forMatched text
description: "...ront charge, but it provisions live SIP credentials, and any traffic you then send over the..."
The description mentions "credentials".
What this check looks forMatched text
description: "...atus, per-unit USD pricing and your SIP credentials (username + password) for each. Require..."
History
Changes to its tool definitions and to its remote endpoint, newest first.
PostureRemote posture changed. Tools that change things, listed without credentials: 20 -> 31; RFC 9728 protected resource metadata: none -> https://packetexchange.io/.well-known/oauth-protected-resource/mcp/http; Authorization servers: none -> ht…View diff
- Tools that change things, listed without credentials: 20 to 31
- RFC 9728 protected resource metadata: none to https://packetexchange.io/.well-known/oauth-protected-resource/mcp/http
- Authorization servers: none to https://packetexchange.io
- Dynamic client registration: not observed to yes
- Client ID metadata documents: not observed to no
Tools5 changed.~5 changedView diff
- Changed:switch_assign_customer_routingDeletes or cancelsinputSchema
- Changed:switch_bulk_upsert_supplier_ratesDeletes or cancelsinputSchema
- Changed:switch_run_nettingChanges thingsinputSchema
- Changed:switch_set_route_plan_entriesDeletes or cancelsinputSchema
- Changed:wmmn_route_detailsOnly readsinputSchema
Install
Pinned to version 2.1.0. Values in angle brackets are yours to fill in.
claude mcp add --transport http packetexchange https://packetexchange.io/mcp/http --header 'Authorization: <your-authorization>'claude mcp add --transport http packetexchange https://packetexchange.io/mcp/http --header 'Authorization: <your-authorization>'Run in a terminal.
Claude Code permissions
Allow the reads, ask before a change or a way out, deny the deletes. Paste into .claude/settings.json.
{
"permissions": {
"allow": [
"mcp__packetexchange__wmmn_list_routes",
"mcp__packetexchange__wmmn_route_details",
"mcp__packetexchange__wmmn_market_summary",
"mcp__packetexchange__wmmn_resolve_route",
"mcp__packetexchange__wmmn_list_purchases",
"mcp__packetexchange__wmmn_get_call",
"mcp__packetexchange__wmmn_get_balance",
"mcp__packetexchange__wmmn_list_transactions",
"mcp__packetexchange__switch_preview_routing",
"mcp__packetexchange__switch_list_customers",
"mcp__packetexchange__switch_list_suppliers",
"mcp__packetexchange__switch_list_invoices",
"mcp__packetexchange__switch_fraud_events"
],
"ask": [
"mcp__packetexchange__wmmn_purchase_route",
"mcp__packetexchange__wmmn_send_sms",
"mcp__packetexchange__wmmn_make_call",
"mcp__packetexchange__wmmn_send_voice_otp",
"mcp__packetexchange__wmmn_verify_start",
"mcp__packetexchange__wmmn_verify_check",
"mcp__packetexchange__switch_analytics_overview",
"mcp__packetexchange__switch_analytics_breakdown",
"mcp__packetexchange__switch_create_customer",
"mcp__packetexchange__switch_create_supplier",
"mcp__packetexchange__switch_create_invoice",
"mcp__packetexchange__switch_ar_aging",
"mcp__packetexchange__switch_create_payable",
"mcp__packetexchange__switch_run_netting"
],
"deny": [
"mcp__packetexchange__wmmn_topup_x402",
"mcp__packetexchange__switch_set_route_plan_entries",
"mcp__packetexchange__switch_assign_customer_routing",
"mcp__packetexchange__switch_bulk_upsert_supplier_rates",
"mcp__packetexchange__switch_bulk_upsert_sell_rates",
"mcp__packetexchange__switch_update_fraud_settings"
]
}
}Tools that only read are allowed, tools that change things ask first, tools that delete or cancel are denied. A read that can send data out (it takes a URL, say) asks first too. It assumes the server is installed as packetexchange. Review it before you use it.
Capability tags behind 8 rules
- mcp__packetexchange__wmmn_send_sms untrusted-input, outbound
- mcp__packetexchange__wmmn_make_call untrusted-input, outbound
- mcp__packetexchange__wmmn_get_call untrusted-input
- mcp__packetexchange__wmmn_send_voice_otp untrusted-input, outbound
- mcp__packetexchange__wmmn_verify_start untrusted-input, outbound
- mcp__packetexchange__wmmn_topup_x402 untrusted-input
- mcp__packetexchange__switch_analytics_overview outbound
- mcp__packetexchange__switch_analytics_breakdown outbound
Pin
Keep the decision: fail CI when the tools change, follow every change, or show a badge.
Pin in CI
Fail a build when the tools you reviewed are not the tools the server serves today.
curl -s "https://radar.anlyon.com/api/v1/servers/io.packetexchange/mcp/verify?hash=651939062143f2b4405a2e13f9d8434fdc0331942d0ef21975b9cfb104915a3c" | jq -e '.changed == false'Pinned hash 651939062143, from Nov 17, 2026, 18:17 UTC.
Watch
Get told when its tools change: a new tool, a new description, a new class.
https://radar.anlyon.com/servers/io.packetexchange/mcp/changes.xmlAn Atom feed. Works in any feed reader or Slack's RSS app.
Badge
[](https://radar.anlyon.com/servers/io.packetexchange/mcp)Source: MCP Registry · repository. Synced Oct 21, 2026, 18:52 UTC.
PacketExchange MCP server: questions and answers
Answered from the scan above, not from the publisher's description.
Is the PacketExchange MCP server safe?
- No scan can promise that. What its definitions show: 33 tools, 20 of which can change, send or delete something, and 5 findings from fixed checks. Allow the reads, ask before each change, and deny the deletes; the permissions block on this page does that.
Can the PacketExchange MCP server delete data?
- Yes. 6 tools can delete, cancel or revoke something: wmmn_topup_x402, switch_set_route_plan_entries, switch_assign_customer_routing, switch_bulk_upsert_supplier_rates, switch_bulk_upsert_sell_rates and switch_update_fraud_settings.
Can the PacketExchange MCP server do things that cannot be undone?
- Yes. 8 tools can send, pay, publish or merge: wmmn_purchase_route, wmmn_send_sms, wmmn_make_call, wmmn_send_voice_otp, wmmn_verify_start, switch_analytics_overview and 2 more.
What tools does the PacketExchange MCP server have?
- 33 tools: 6 destructive, 8 irreversible, 6 write, 13 read. wmmn_topup_x402, switch_set_route_plan_entries, switch_assign_customer_routing, switch_bulk_upsert_supplier_rates, switch_bulk_upsert_sell_rates, switch_update_fraud_settings, wmmn_purchase_route, wmmn_send_sms and 25 more.
Does the PacketExchange MCP server need sign-in or OAuth?
- Its remote endpoint listed its tools to an anonymous client, without sign-in.
Does the PacketExchange MCP server send data to outside services?
- 6 tools can carry data out (it sends, posts or publishes, or takes a recipient, channel or URL): wmmn_send_sms, wmmn_make_call, wmmn_send_voice_otp, wmmn_verify_start, switch_analytics_overview and switch_analytics_breakdown.
Has the PacketExchange MCP server changed its tools?
- Yes, once since Anlyon Radar started tracking it, most recently on Nov 17, 2026. Every change is on this page and in its Atom feed.