[ AI Runtime Infrastructure ]

Run AI agents on
secure cloud runtimes.

e2a provides isolated compute, browser, and desktop environments for AI workloads that need reliable infrastructure, scoped credentials, and production-grade controls.

MicroVM isolation · Tenant-scoped networking · Usage-based capacity
runtime plan — e2a
< 500msBoot time
3 runtimesProfiles
MicroVMIsolation

Capabilities

Everything your AI workload needs
to run with control

Fast Runtime Launch

< 500ms boot

Provision isolated cloud runtimes for AI workloads with predictable startup, resource limits, and lifecycle controls.

Long-Running Workloads

Keep workflows running for minutes or hours with durable state, health checks, logs, and completion callbacks.

Realtime Control Plane

Stream runtime output, send input, and observe state transitions through a managed interactive channel.

Scoped Secret Handling

Inject provider credentials and application secrets only where they are needed, without storing them in shared logs.

MicroVM Isolation

AWS-grade

Run each workload behind a hardware virtualization boundary with tenant separation and controlled network policy.

Credit-Based Billing

Pay only for CPU, RAM, and duration consumed. No seat licenses, no minimums, no surprise bills.

Platform-First

Built for AI workloads.
Operated like infrastructure.

Give agentic systems a governed place to run: isolated compute, interactive surfaces, scoped secrets, lifecycle events, and usage visibility from one control plane.

  • Provision runtimes with clear resource limits
  • Stream output and lifecycle events in real time
  • Connect from backend services or orchestration systems
  • Keep model provider credentials under your control
runtime:
  profile: standard
  isolation: microvm
  surfaces:
    - shell
    - browser
  limits:
    cpu: 2
    memory: 4Gi

Runtime Profiles

Choose the right surface
for your workload

bash — e2a terminal
user@rt:~$ ls /workspace
data.csv process.py output/
user@rt:~$ python process.py
Processing 12,847 rows...
✓ Done — result.json written

Isolated Linux workspace

Runtime Shell

Run developer tools and data workflows in a clean Linux workspace with bounded CPU, memory, and storage.

bashpythonnodeapt-get
Use Template →
https://research.example.com
↳ page review completed · screenshot captured

Managed Chromium

Browser Runtime

Run controlled browser sessions for workflow testing, public page review, and screenshot capture.

playwrightchromiumDOMscreenshots
Use Template →
virtual desktop — 1280×720
target area
↳ desktop control active · input routed securely

Graphical workspace

Desktop Runtime

Use a full graphical workspace for applications that need screen, mouse, and keyboard access.

VNCxdotoolOCRfull-desktop
Use Template →

Pricing

Only pay for what your
runtimes consume

Credit-based pricing. No minimums, no seat licenses. Predictable costs at any scale.

Credit Rates

CPU
1.0 creditsper vCPU-minute
RAM
0.5 creditsper GiB-minute
Model Input
1.0 creditsper 1K tokens
Model Output
3.0 creditsper 1K tokens
Runtime Operations
10.0 creditsper call
Workspace Storage
0.1 creditsper MiB-month

Free

100 credits on signup

No credit card required. Start building immediately.

  • All 3 runtime profiles
  • Realtime output streaming
  • Scoped secret injection
  • Community support
Get Started Free
Most Popular

Pay As You Go

Buy credits via Stripe

Top up your balance whenever you need more. No commitments.

  • Everything in Free
  • Webhook callbacks
  • Priority boot queue
  • Email support
Buy Credits

Auto-Topup

Never run out

Set a threshold and top-up amount. We charge automatically when your balance is low.

  • Everything in Pay As You Go
  • Automatic credit refills
  • Configurable threshold + amount
  • Priority support
Sign Up Free

Developer Endpoints

API entry point for this environment.

Use these base URLs for the current environment. Account access, API keys, and detailed integration contracts are provided through the dashboard during onboarding.

Control API

https://api.sandbox.deictic.ai

Authentication, runtime lifecycle, secrets, usage, billing, and account operations.

Realtime Runtime

wss://rt.sandbox.deictic.ai

Interactive runtime channel used after your service creates a runtime session.

Quick Start

Start with a Controlled Pilot

Validate the runtime model with one workflow, then expand capacity as your AI workload grows.

1

Choose a runtime profile

Select shell, browser, or desktop capacity based on the workload surface your AI system needs.

2

Connect your control plane

Route work from your service into isolated cloud runtimes with scoped credentials and lifecycle events.

3

Scale with usage-based capacity

Move from a pilot to production with metered CPU, memory, storage, and runtime duration.

Start Pilot →