SWITCH
SWITCH is an agentic AI platform for urban fleet and logistics operators. Its Myles agent answers operational questions, forecasts demand, plans rebalancing and simulates network decisions across 3,800+ cities. Pro access starts at 29 USD per month.
What is SWITCH?
SWITCH — a brand that plays on “Street WITCHer” — is an agentic AI platform for urban mobility and logistics, published by Switch SRL in Rome. Matteo Forte (CEO) and Nunzio Gianfelice founded it around a car-sharing problem they knew at first hand: vehicles sitting idle on one side of the city while demand climbs on the other. The flagship product is Myles, a conversational agent presented as the AI teammate for fleet and logistics operators. It draws on two claimed data layers: Context Intelligence — 3,800+ cities, 350 million modelled trips a month, weather and seasonality — and Fleet Intelligence, the customer’s own operating history. Four jobs are covered: analytics, forecasting, operations and planning. Answers come back in three forms: text with a chart, exportable documents, or actions the agent carries out itself, on demand or on a schedule. Two proprietary engines sit underneath — Pulse-AI for forecasting and simulation, OptiMesh for routing — and the technology page claims forecast accuracy of up to 98%. The pitch against general-purpose assistants is blunt: “A general model has read everything about mobility. It has never been on the road.” Two Enterprise modules extend the platform: Urbiverse, an agent-based simulation twin, and Urban CoPilot, real-time execution with a mobile interface for field crews. A REST API on api.getswitch.io exposes analytics, forecasting, operations, routing, fleet management, vehicle integrations and synthetic trip generation in JSON and GeoJSON, and a Myles MCP server opens the same capabilities to agent-to-agent access. The intended buyers are shared-mobility and rental operators, last-mile and 3PL logistics teams, vehicle manufacturers, local authorities and mobility consultancies. The company is backed by EIT Urban Mobility and was accelerated by Berkeley Skydeck Europe Milano. One caveat matters for anyone reading this today: as of 24 August 2026 Myles is still on a waiting list, while the Enterprise platform is already deployed, with named case studies covering Wayla, Elerent, micromobility hubs in Turin and the 2026 Milan winter games. The claim SWITCH puts forward — thirty seconds or less from question to answer — is therefore observable today mainly on the Enterprise side.
What it does
- Ask a plain-language question about your operation and get the answer back with the chart.
- Forecast demand zone by zone and hour by hour, with confidence bands.
- Generate a rebalancing plan and assign it to field crews.
- Optimise a multi-stop route and recompute it in real time.
- Simulate a city launch, a fleet resizing or a depot location before committing.
- Produce exportable documents: performance reports, demand analyses, tender packs.
- Schedule the agent to run on its own, every morning or every shift.
When to use SWITCH / When not to
A quick filter to help you decide if SWITCH is the right fit.
When to use SWITCH
- Shared mobility operators — scooters, bikes, cars — fighting supply and demand imbalance that shifts zone by zone and hour by hour.
- Last-mile delivery teams and 3PL providers that need multi-stop routing and clear prioritisation of their rounds.
- Vehicle rental companies and vehicle manufacturers, both addressed by dedicated solution pages.
- Local authorities and mobility consultancies that want to test a capital decision before committing to it — a new city launch, a fleet resizing, a depot or charging-point location — through the Urbiverse simulation module.
- Operations teams with no data scientist on staff: the interface is plain language, with, in the publisher’s words, “No query language, no dashboard”, and you can start on sample data from 3,800+ cities instead of an integration project.
When not to use SWITCH
- Anyone hoping to sign up for Myles today: as of 24 August 2026 the self-service route ends at a waiting-list form, not a checkout.
- Developers who need programmatic access: API keys ship with Myles Enterprise only, and there is no SDK — “Is there an SDK? Not today.”
- Pro subscribers expecting advanced optimisation or simulation: Urbiverse and Urban CoPilot are marked Enterprise only, and Pro is capped at 50+ cities and 25 airports against the full 3,800+ city catalogue.
- Teams working outside mobility and logistics: the Context Intelligence layer is built on trip data and has nothing to bite on elsewhere.
- Buyers who want simulation as a monthly SaaS line item, or field crews expecting a downloadable app: Urbiverse is sold as a project engagement, and no App Store or Google Play listing exists — the mobile interface belongs to Urban CoPilot, on the Enterprise side.
How to use SWITCH
A typical end-to-end flow, from setup to results.
- Pick your entry point: Pro sign-up currently routes to the Myles waiting list, while Enterprise prospects book a 20-minute call with the technical team.
- STEP 01 — Connect: plug in your feeds, or start on sample data from 3,800+ cities. The publisher promises “minutes to get going, not an integration project”.
- On Enterprise, connect fleet, telematics and booking platforms through secured APIs — the stated principle is “no rip and replace”.
- STEP 02 — Ask: type the question the way you would put it to your head of operations. No query language, no dashboard to learn.
- STEP 03 — Act: read the answer and the next action it suggests, then let Myles execute it.
- STEP 04 — Automate and improve: schedule runs every morning or every shift; the agent builds its own skills around your operation.
- Collect the artifacts — performance reports, demand analyses, tender packs — from your workspace, then export or share them; a demonstration video on the Myles page shows what they look like.
- For developers: provision API keys with a scope and an optional expiry, or use OAuth, sending a bearer token per request — or connect an AI agent through the Myles MCP, with no keys to manage.
- For an Urbiverse project: load historical data, define the scenario, run the comparison, then collect the quantified output.
Pros & Cons
Pros
- Genuinely vertical: the models are trained on movement data rather than on text, behind two proprietary engines, Pulse-AI and OptiMesh.
- No integration project to start: sample data from 3,800+ cities is available from day one.
- Natural-language interface with no query language and no dashboard to learn, and an agent that grades its own certainty through confidence bands on forecasts.
- A complete chain from question to answer to plan to execution, with the same intelligence available at several levels of control.
- Outputs are exportable, shareable artifacts rather than screens you have to screenshot.
- Low entry point at 29 USD a month with no commitment, cancellable from the dashboard, a 14-day free trial with no charge if cancelled in time, and a Pro to Enterprise move that keeps data, integrations and configuration.
- European publisher with its data controller in Italy, backed by EIT Urban Mobility, with named deployments carrying published figures (Wayla at 92% forecast accuracy, Elerent at +25% fleet performance) and agent-to-agent access through MCP without key management.
Cons
- Myles is not available in self-service as of 24 August 2026: the “Start Free” button leads to a waiting list.
- There is no permanent free plan, only a 14-day trial.
- The pricing page contradicts itself: 29 USD a month on the plan card, 30 USD a month in two of its own FAQ answers.
- The API and the advanced modules — Urbiverse, Urban CoPilot, simulation, live integrations — are Enterprise-only, and Enterprise pricing is not public; there is no SDK either, only direct REST calls.
- No terms of service are published, no DPA is offered, no processor list is published and no SOC 2 or ISO 27001 certification is displayed.
- No hosting country or region is stated for customer data, and no support email address is published — contact goes through a form or the general address.
- The product FAQ says customer data never trains shared models, while the 2025 funding release says processed trips further trained the startup’s models; elsewhere on the site, “250+ Integrations” is claimed but barely fifteen logos are shown, without alt text, next to an unfilled placeholder (“Typical engagements run [X-X weeks]”) and a typo (“SWITCH IINTEGRATION”).
Pricing & Plans
There is no permanent free plan. The lowest paid entry point is the Pro plan at 29.00 USD per month, billed monthly, cancellable at any time from the dashboard at no charge, and preceded by a 14-day free trial that converts automatically into a paid subscription, with a reminder sent three days beforehand. It should be noted that the FAQ on the same pricing page quotes that conversion at 30 USD per month, and annual billing at 330 USD per year with one month free — a discrepancy the publisher does not explain. Pro carries unlimited questions and an unlimited number of vehicles, with no per-vehicle cap and no usage fees, and credits may be topped up at any time. Enterprise is quoted on request, most often as an annual contract with per-vehicle and per-city components, depending on fleet size, number of cities and modules. Urbiverse is sold as a project engagement rather than a monthly SaaS subscription, scoped by the number of scenarios, the number of cities and complexity, while Urban CoPilot taken standalone is priced per vehicle per month plus per city per month.
- the four jobs plus autonomy
- unlimited questions and vehicles
- 50+ cities and 25 airports
- credit top-ups
- a 14-day free trial
- and 330 USD on annual billing.
- full integrations
- custom models
- optimisation and simulation
- the complete 3
- 800+ city catalogue plus API access
- guided onboarding
- SSO and SLA.
- fleet sizing
- infrastructure siting
- city launch
- regulatory impact analysis and electrification modelling.
- autonomous generation of rebalancing tasks
- crew assignment
- OptiMesh routing
- battery and EV logistics
- multi-stop tasks and a field mobile interface.
Data, GDPR & hosting
A consolidated view of how SWITCH handles your data.
GDPR overview
SWITCH Srl is established in Rome, inside the European Union, so no Article 27 representative is required. The privacy policy hosted on iubenda, last updated on 29 July 2024, carries a dedicated section on the rights of users under the GDPR: withdrawal of consent, objection, access, rectification, restriction, erasure, portability and the right to lodge a complaint with a supervisory authority. Legal bases are enumerated — consent, performance of a contract, legal obligation, public interest task, legitimate interest — and transfers outside the user’s own country are declared possible, with Ireland and the United States named as processing places for third-party services. The gaps are equally explicit: no data protection officer is named, no DPA is offered or mentioned, no SOC 2 or ISO 27001 certification is displayed, and the words “GDPR compliant” appear nowhere. Compliance has to be inferred from the published apparatus, never read as a claim.
Who owns the data?
SWITCH Srl, Via Giacomo Peroni 396, 00131 Rome, is named as owner and data controller in the privacy policy hosted on iubenda, with privacy@getswitch.io as the contact address. On the product side, the Myles FAQ is unambiguous about ownership: “Your data is never used to train shared models or shared with anyone. It’s yours.” That statement lives in a FAQ, not in a contract: the published policy covers the website, no terms of service exist, and no separate document covers customer fleet data. Named third parties with access to website data are Google Ireland, Google LLC, LinkedIn Corporation, HubSpot Inc. and InnoCraft; the current processor list is available on request only.
Reuse rights
No terms of service are published, so nothing contractually defines what a customer may do with the data and outputs they take out of the platform — in practice, documents, demand analyses and tender packs are presented as exportable and shareable artifacts saved to the customer’s own workspace. The privacy policy, which covers the website alone, lists analytics, remarketing and behavioural targeting, display of external content, user database management, tag management and user contact as its purposes, through Google Analytics (Universal and GA4), Matomo Cloud, Google Fonts, LinkedIn Website Retargeting, Google Tag Manager and HubSpot CRM. Its US state-law section declares Sold or Shared: Yes and Targeted Advertising: Yes for several data categories. Two first-party sources then pull in opposite directions on the product itself: the Myles FAQ says customer data never trains shared models and is never shared, while the pre-seed announcement of 6 February 2025 says processing over five million trips served to “further training the startup’s AI models”. The publisher has not reconciled the two, and no training opt-out is documented.
Data retention & training
Hosting summary
The site does not say where customer data is hosted: no country, no region, no cloud provider is named. The iubenda privacy policy states only that data is processed at the owner’s operating offices and in any other place where the parties involved in the processing are located, and it explicitly allows transfers to a country other than the user’s own. The only processing places named are those of the website’s third-party services: Ireland (Google, Google Fonts) and the United States (Google Analytics 4, LinkedIn, HubSpot). The publisher itself is established in Rome, therefore inside the European Union, and the declared data controller sits at a Rome address. One element the site does not declare, observed during collection: getswitch.io resolves to a server in Italy (86.107.37.54, AS52030 Server Plan S.r.l., Rome) — which concerns the website, not the fleet data the platform processes. For an operator with its own GDPR obligations, the hosting jurisdiction of operational data therefore remains an open question to put to the vendor, alongside the absence of any data processing agreement.
Things to keep in mind
Risks and trade-offs to weigh before adopting SWITCH.
- Access does not match the promise: “Start Free” and “Start free trial” all lead to a waiting-list form, so budgeting today on the 29 USD plan means budgeting for something you cannot yet buy.
- The advertised price is contradicted on its own page: 29 USD a month on the card, 30 USD a month in the FAQ, 330 USD a year on annual billing.
- Two first-party statements about training disagree: the product FAQ says customer data never trains shared models, while the February 2025 funding release says the trips processed further trained the startup’s models. No training opt-out is documented.
- The published privacy policy covers the website, not the processing of customer fleet data, and no separate contract is published — nor are any terms of service accessible on the site. No DPA is offered, no security certification is displayed, the processor list is available on request only, and no hosting country or region is stated.
- That same policy declares Sold or Shared: Yes and Targeted Advertising: Yes for several data categories under US state laws — an unusual position for a European B2B publisher, and one worth raising directly.
- Several claims cannot be checked from the site: “250+ Integrations” shown as barely fifteen logos without alt text, unsourced performance figures (-81% idle time, +25% efficiency, 67% rebalancing ROI, up to 98% forecast accuracy), an unfilled placeholder in the Urbiverse FAQ (“Typical engagements run [X-X weeks]”), and two different street numbers for the same Rome address (386 in the footer, 396 in the privacy policy).
- Delegating judgement is the quiet risk: the agent generates plans, assigns tasks to crews and can be scheduled to run every morning or every shift. The publisher states that the dispatcher always keeps control and that exceptions are escalated to a human — a safeguard worth actively maintaining, so that operational instinct about your own city does not quietly fade into whatever the schedule produces.
Setup & Integrations
Technical difficulty
Low for Pro, moderate for Enterprise. Pro claims no integration project: start on sample data from 3,800+ cities, or connect by API to your existing platform for trips, vehicle positions and zones — announced in minutes, with no query language or dashboard to learn. Enterprise involves secured API connections to fleet, telematics and booking systems, plus GBFS/GTFS feeds and demand signals, with white-glove onboarding, SSO and SLA. Developers work in direct REST with no SDK, using API keys or OAuth, or through the Myles MCP. Urbiverse needs trip history, fleet composition, zone definitions and operating parameters, in standard exports.
Deployment
Integrations
Behind SWITCH
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement SWITCH.
Frequently asked questions
How is this different from ChatGPT, Gemini, Claude or Perplexity?
What is Context Intelligence?
Does it work with my current fleet management system?
What happens after the 14-day free trial?
Is there a discount on annual billing?
Do I need Enterprise to use the API?
Is there an SDK?
Is my data secure?
What happens if the system gets it wrong?
Can I simulate a city where I have not launched yet?
Should you pick SWITCH?
SWITCH is an unapologetically vertical tool, and that is its strongest argument: the value comes from movement data — 350 million modelled trips a month across 3,800+ cities — and from two proprietary engines, not from a general-purpose model asked politely. The chain it offers, from question to plan to execution, is rare on this market: few products let the same intelligence answer, recommend and then act, with the dispatcher keeping the final say. Three reservations should temper the enthusiasm. First, Myles in self-service is not open as of 24 August 2026 — every Start Free button leads to a waiting list, so what you can actually buy today is the Enterprise platform, not the 29 USD subscription. Second, most advanced capabilities and the whole API sit behind Enterprise, whose price is not public, while Pro is capped at 50+ cities and 25 airports. Third, the contractual and security documentation is thin: no terms of service, no DPA, no published processor list, no security certification, no stated hosting country for fleet data, and a privacy policy that covers the website rather than the customer data the platform processes. On the other side of the ledger, credibility is real: named deployments with published figures, EIT Urban Mobility backing, a 946,000 EUR pre-seed round and the Berkeley Skydeck accelerator. Before signing, put three questions to the publisher: which price applies, 29 or 30 USD, since the pricing page states both; whether your trip data feeds model training, since the product FAQ and the 2025 funding release disagree; and which contract covers your fleet data, since the published policy does not.
- Choosing a selection results in a full page refresh.
- Opens in a new window.