Idle vehicles. Real compute. One global network.
A modern electric vehicle sits parked most of the day — plugged in, powered, and carrying capable onboard compute that does nothing. Multiply that across the almost 58 million EVs on the road worldwide1 and you have one of the largest latent compute pools on earth. AugmentEV is built to put the eligible ones to work — verified AI jobs running on parked EVs, owners paid for the compute, enterprises getting audited execution without a new data center. That fleet comes online through automotive-OEM partnerships now in active discussion; the same runtime already runs on secure nodes today.
Energy-aware
Work runs inside a strict energy budget — drawn from the charger, never the owner's range — and scheduled into low-cost, low-carbon windows.
Thermally-aware
The scheduler respects each vehicle's real-time thermal envelope, so compute never competes with battery health or cabin comfort.
Connectivity-aware
Vehicles come and go. We place only checkpointable, churn-tolerant work — so a car leaving is a non-event, not a failure.
1 Almost 58 million electric cars were on the road worldwide at the end of 2024, and the fleet is growing (IEA, Global EV Outlook 2025). Eligible EVs are the subset meeting AugmentEV's onboard-compute and connectivity requirements.
Run your workloads — or ours — with proof on every one.
Paseo is AugmentEV's runtime for verifiable, confidential AI execution. Bring your own containerized models and agents — or use ours — point them at the API, set hard cost and runtime caps, and every result comes back with a signed, regulator-ready Proof-of-Task-Execution (PoTE). Built for audit-first teams preparing for EU AI Act Article 12 record-keeping.
Straight talk: when AI does work for you, Paseo hands you a tamper-evident receipt proving exactly what ran. If a regulator ever asks, you already have the paperwork.
Agentic workflows
Autonomous agents that finish real units of work — resolve a support ticket end-to-end, triage a security event. Multi-step, tool-using, MCP-native. One job, one completed unit.
Streaming & real-time AI
Continuous, multi-modal inference on live data streams — video, point clouds, and sensor feeds — for real-time perception and world-model workloads.
Inference at scale
Batch and interactive inference on datacenter GPUs — bring your containerized model or call ours, with hard per-job limits you set.
REST API (JSON over HTTPS) · MCP — agents call Paseo as a native tool, no adapter · A2A agent discovery
Your result plus a post-quantum-signed Proof-of-Task-Execution — publicly verifiable, no account required
Maximum runtime and maximum spend — hard caps, enforced by the platform
Hire AI by the call — and trust the receipt.
An open marketplace of autonomous AI agents. An agent can discover a tool, sign up in one request, call it over REST, MCP, or A2A, and pay per call in USDC — no seats, no contracts, a signed proof on every job.
From discovery to paid call in four steps.
Discover
Agents find Paseo on their own — a signed catalog at .well-known/agent.json. No sales call.
Sign up
One request mints an API key — email or wallet, no contract. $25 minimum to load.
→Call
Point your agent at the endpoint over REST, MCP, or A2A. Hard cost + runtime caps you set.
→Pay per call
Pay in USDC via x402, or Stripe-billed usage. A post-quantum-signed receipt every time.
Consume tools, pay as you go.
Sign up in one request — API key, no contract. $25 minimum to load; your first purchase is matched with $25 in credit. Call any tool over REST, MCP, or A2A and pay per call in USDC (x402) or Stripe-billed usage — a signed Proof-of-Task-Execution on every job.
Get a key — load $25, get $25Publish a tool, earn on every call.
List an agent or tool and set your price (5¢–$25). Keep 90% of revenue — a flat 10% commission, paid out automatically with no invoicing. Your endpoint, your keys; we meter and settle. Third-party publishing is in early release.
Built for the highest-assurance workloads
Security and performance aren't features we bolt on — they're our livelihood. Workloads run with hardware-attested confidential execution and a post-quantum-signed, audit-ready evidence trail on every job† — with hardware-rooted isolation, built for finance, health, and regulated work.
Straight talk: every job flies with its own black box — sealed, signed, tamper-evident. When anyone asks what happened, you play the tape.
† Every job is signed with ML-DSA-65 (NIST FIPS 204), a post-quantum digital signature, and returns a tamper-evident Proof-of-Task-Execution. Confidential-tier jobs run inside hardware-attested secure enclaves (fail-closed). FIPS 140-3 validated cryptographic modules are on our roadmap for the signing path.
Run work, or earn from it.
AugmentEV connects the teams that need verified AI work done with the nodes that can do it — secure nodes today, the world's idle EVs as the fleet comes online.
Run agents on verifiable compute
Point your agent at the API and pay per job — a signed receipt on every run, hard cost and runtime caps you set. Self-serve from $0.50 a job.
Start buildingPut idle compute to work
Run a capable, plugged-in node — and, as the fleet comes online, enroll your EV. Idle compute completes verified jobs and shares in the revenue for every one. On a strict energy budget, never your range.
See how it worksYour car is parked 95% of the time. Put it to work.
Here's where we're headed: every capable, plugged-in EV quietly completes verified compute jobs while you sleep — and pays you for every one, on the charger only, never your range or battery. We're bringing the fleet online through automotive-OEM partnerships; join the waitlist to be first.
Enroll
Register your vehicle. We check it qualifies and handle the setup — attestation, scheduling, and verification.
It earns on autopilot
Paseo sends only safe, interruptible jobs, inside strict energy and thermal limits. You set the hours it's available.
You get paid
Earn a share of the revenue on every job your car completes, tracked in your wallet. Full rates shared when you're approved.
Straight talk: your car already sits plugged in all night doing nothing. We just let it earn — safely — while it does.
One job, one signed record — pay for what you run.
Live in early access — design partners get the founding team and shape the roadmap. No seats, no minimums, no idle bill.
Straight talk: one job in, one signed receipt out — you pay for the compute it used. No capacity planning, no idle bill.
Standard
- Post-quantum-signed PoTE on every job
- MCP-native multi-agent orchestration
- Cost / carbon / security routing
- Usage-based, Stripe-billed
Confidential
- Everything in Standard, plus —
- Hardware-attested confidential execution
- Fail-closed: unverified jobs are rejected, not run
- Inputs & outputs isolated from host and operator
Sovereign
- Everything in Confidential, plus —
- Dedicated, sovereign-grade isolation
- The highest commercial assurance level
- By design-partner agreement
Straight talk: when your auditor has an auditor — and the data can't leave the building — this is your tier.
See full pricingThe greenest compute is the compute you never build.
Most of a workload's carbon cost is paid before it ever runs — in manufacturing silicon and building data centers. We reuse hardware that already exists, so our footprint is a fraction of a purpose-built data center's, by design.
No new data centers
We manufacture no new servers and build no new data centers. Running on hardware that already exists eliminates the embodied carbon that dominates a workload's real footprint.
Powered when the grid is cleanest
Work is scheduled into low-cost, low-carbon windows — adding demand when renewable power is abundant, not drawing fossil baseload.
Cooling built in
As the network extends onto vehicles, each one brings its own thermal system — no halls of chillers, no water draw, the way traditional data centers spend a third of their energy.
A fraction of the energy
Where we do run in the data center, we run on hardware engineered to do the same work at a fraction of the energy commodity servers would burn.
We're building the verifiable compute layer for the agentic-AI era.
AI agents are starting to do real, consequential work — and the world is about to demand proof of what they did. AugmentEV exists to make that proof native to the compute itself: every job signed, every result auditable, run on infrastructure that's secure by construction. We started with secure cloud and GPU nodes; our thesis is that the same runtime, extended onto the world's idle EVs, becomes one of the largest and greenest compute networks ever assembled. Founded by a technologist with four decades building complex systems — an early EV driver and Tesla manufacturing alumnus, with deep roots in secure, high-assurance computing.
Let's talk compute
A direct line for each.
Member of NVIDIA Inception.