FC Agent Sandbox Plan Features by Tier

Updated at:

Multi-tier Subscription Plans: Target Customers & Scenarios

Plan

Target Customers

Typical Scenarios

Compute Stability & Performance

Eco

AI Startups、
Individual Developers

Agent Exploration & Functional Testing:
Single-step debugging, Tool Use chain verification, low-frequency asynchronous long-running tasks.

Lowest Barrier to Entry:
Cost-effective choice; allows for minor compute fluctuations.

Std(Recommended)

Mid-market Enterprises
AI SaaS Providers

Enterprise Copilots & Productivity Tools:
Daily office assistants, data analysis (Code Interpreter), document parsing & format conversion.
Large-scale Offline Sampling & Parallel Training:
Cost-effective for high-volume tasks; tolerates second-level cold starts; provides massive low-cost sandboxes.


Balanced & General-purpose:
Stable performance with low-latency response; suitable for enterprise-grade production workloads.

Pro(Recommended)

Top-tier AI Application Enterprises, MaaS/Agent Platforms

High-concurrency C-side Agent Apps:
Multi-agent collaboration, code execution for massive concurrent users, and heavy complex computations.
Large-scale Online RL Simulation:
Instant provisioning of massive sandbox environments; supports ultra-low latency, high concurrency, and high-throughput RL environment sampling & simulation.


Elastic High Throughput:
Supports million-level sandbox concurrency with ultra-low latency.
Ideal for high-concurrency Agent code execution and large-scale distributed high-throughput RL sampling.

Plan Features

Category

Feature

Eco

Std (Recommended)

Pro (Recommended)

Performance & Scale

Compute Quality

Occasional compute fluctuations

Stable performance with low-latency response

High elasticity and throughput, ultra-low latency

Cold Start Performance per Sandbox Instance (without pre-warming)

Hundreds of milliseconds

Hundreds of milliseconds

Hundreds of milliseconds

Warm Start Performance after Sandbox Resource Pool Pre-warming (paid feature)

unsupport

supportPre-warming reduces startup time to tens of milliseconds

supportPre-warming reduces startup time to tens of milliseconds

Deep Hibernation Wake-up Efficiency under High Concurrency *

unsupport

Hundreds of milliseconds → 3 seconds

Hundreds of milliseconds → 3 seconds

Light Hibernation Wake-up Efficiency under High Concurrency

unsupport

unsupport

1 millisecond

Hibernation/Resume Capabilities

Deep Hibernation *

unsupport

support

support

Light Hibernation *

unsupport

unsupport

support

Maximum Lifetime

Maximum Sandbox Instance Lifetime

24 hours

24 hours

7 days

Maximum Retention Duration in Deep Sleep State

unsupport

7 days

30 days

Dynamic Mounting

Dynamic Mounting of Cloud Storage Products on the Platform Side

support

support

support

Security

L7 Network Traffic Control (IP + Domain)

support

support

support

Operations & Maintenance

Logging and Observability

support

support

support

  • Notes:
    1.Deep Hibernation Resume Efficiency under High Concurrency: Refers to the total time required to restore Deep Hibernation instances and rebuild network connections in a high-concurrency scenario (10,000+ concurrent requests) with a 2C4G instance specification.
    2.Deep Hibernation: Comparable to E2B's Pause/Resume functionality. The maximum lifetime for an active (unpaused) sandbox is 24 hours for Eco/Std plans and 7 days for the Pro plan. The timer automatically resets upon pausing, granting the full duration again upon resumption.By cycling through "pause-resume-pause-resume," the sandbox's online duration can be continuously extended.
    3.Light Hibernation: Exclusive to the Pro plan. Leveraging request-awareness and memory snapshot technologies, instances automatically enter shallow sleep when idle (CPU usage < 0.01 core / Network traffic < 1,000 Bytes/s / No incoming requests) to reduce costs (No vCPU charges are incurred during shallow sleep; memory and disk storage continue to be billed according to standard rates). Instances are woken up ultra-fast within 1 ms upon receiving new requests.


Feature differences across subscription plans will continue to evolve. Stay tuned!

Subscription Plan Pricing

For more details, please refer to the Pay-as-you-go documentation.

Subscription Plan Management

Purchase & Billing:No activation required; billed on a pay-as-you-go basis. Set the subscription plan when creating or managing a Team, with options for Eco, Std, or Pro tiers.
Activation Mechanism: Subscription plan attributes are inherited from the Team level downwards. All Templates and Sandboxes under the current Team automatically synchronize with these plan attributes, ensuring global consistency.
Changes & Management:

  • Upgrade: Supports one-way seamless upgrades (Eco → Std → Pro), which can be performed directly in the Team settings.

  • Downgrade: To prevent service disruption caused by feature reduction for existing workloads, online downgrades are currently not supported. If you need to downgrade, please create a new Team or contact technical support (DingTalk Group: 179855020297).