Compute API (Modal Sandbox)

Isolated code execution in secure cloud containers via Modal. Create sandboxes, execute arbitrary commands, manage lifecycle. Supports resource limits, network isolation, and automatic cleanup. Designed as a secure runtime for AI agent code execution.

Base URL: https://api.jarvisclaw.ai/v1/marketplace/compute

Endpoints

Method	Endpoint	Description	Price
POST	`/sandbox/create`	Create a new isolated sandbox	$0.01 (CPU, ≤300s)
POST	`/sandbox/exec`	Execute a command in a sandbox	$0.001/request
POST	`/sandbox/status`	Get sandbox status	$0.001/request
POST	`/sandbox/terminate`	Terminate and cleanup	$0.001/request

Pricing

Two billing modes selected automatically by the requested timeout: short bursts pay a flat rate, long-lived sandboxes pay per hour.

Burst — timeout ≤ 300s

Flat rate per sandbox/create, covers the full short-lived session. Lifecycle ops (exec, status, terminate) at $0.001/request.

Config	Price	Resources
CPU	$0.01/sandbox	Up to 1 vCPU, 1 GiB RAM, 300s lifetime
GPU — T4	$0.05/sandbox	Up to 8 vCPU, 32 GiB RAM, 300s lifetime
GPU — L4	$0.08/sandbox	Up to 8 vCPU, 32 GiB RAM, 300s lifetime
GPU — A10G	$0.10/sandbox	Up to 8 vCPU, 32 GiB RAM, 300s lifetime
GPU — A100	$0.20/sandbox	Up to 8 vCPU, 32 GiB RAM, 300s lifetime
GPU — H100	$0.40/sandbox	Up to 8 vCPU, 32 GiB RAM, 300s lifetime

Long-lived — timeout > 300s, up to 24h

Per-hour rate × requested duration, billed upfront in one x402 settlement. No refund on early terminate (same as cloud-instance norms). Hours are exact — 90min on T4 = 1.5 × $1.50 = $2.25.

Config	Per hour	Use cases
CPU (1 vCPU / 1 GiB)	$0.10	Batch jobs, scrapers, schedulers
T4 (up to 8 vCPU / 32 GiB)	$1.50	Light inference, small models
L4 (up to 8 vCPU / 32 GiB)	$2.00	Mid-tier inference
A10G (up to 8 vCPU / 32 GiB)	$2.50	Stable diffusion, 7B inference
A100 (up to 8 vCPU / 32 GiB)	$4.00	Training, 13B–70B inference
H100 (up to 8 vCPU / 32 GiB)	$8.00	Frontier training, low-latency 70B+

POST /sandbox/create

Create a new isolated sandbox container.

Parameters

Parameter	Type	Required	Description
`image`	string	No	Docker image. Default: `python:3.11`
`timeout`	integer	No	Max sandbox lifetime in seconds. ≤300 = flat rate, >300 = per-hour billing. Default: `300`, Max: `86400` (24h)

Request

json

{
  "image": "python:3.11",
  "timeout": 300
}

Response

json

{
  "sandbox_id": "sb_a1b2c3d4e5",
  "status": "running",
  "created_at": "2026-06-13T10:00:00Z",
  "expires_at": "2026-06-13T10:05:00Z"
}

POST /sandbox/exec

Run a command inside an existing sandbox.

Parameters

Parameter	Type	Required	Description
`sandbox_id`	string	Yes	Sandbox identifier
`command`	string or array	Yes	Shell command (string) or command array (e.g. `["python", "-c", "print(1)"]`)

Request

json

{
  "sandbox_id": "sb_a1b2c3d4e5",
  "command": ["python", "-c", "import sys; print(sys.version)"]
}

Response

json

{
  "exit_code": 0,
  "stdout": "3.11.9 (main, Apr  2 2024, 08:25:04) [GCC 12.2.0]\n",
  "stderr": "",
  "duration_ms": 45
}

POST /sandbox/status

Get current status and metadata of a sandbox.

Parameters

Parameter	Type	Required	Description
`sandbox_id`	string	Yes	Sandbox identifier

Request

json

{
  "sandbox_id": "sb_a1b2c3d4e5"
}

Response

json

{
  "sandbox_id": "sb_a1b2c3d4e5",
  "status": "running",
  "created_at": "2026-06-13T10:00:00Z",
  "expires_at": "2026-06-13T10:05:00Z",
  "commands_executed": 3
}

POST /sandbox/terminate

Immediately terminate a sandbox and release all resources.

Parameters

Parameter	Type	Required	Description
`sandbox_id`	string	Yes	Sandbox identifier

Request

json

{
  "sandbox_id": "sb_a1b2c3d4e5"
}

Response

json

{
  "sandbox_id": "sb_a1b2c3d4e5",
  "status": "terminated"
}

Errors

HTTP Status	Code	Description
404	`sandbox_not_found`	The specified sandbox_id does not exist
410	`sandbox_terminated`	The sandbox has already been terminated
408	`sandbox_timeout`	The sandbox exceeded its timeout and was auto-terminated
500	`exec_failed`	Command execution failed due to an internal error

Code Examples

bash

# Create a sandbox
curl -X POST https://api.jarvisclaw.ai/v1/marketplace/compute/sandbox/create \
  -H "Authorization: Bearer sk-your-api-key" \
  -H "Content-Type: application/json" \
  -d '{"timeout": 300}'

# Execute a command
curl -X POST https://api.jarvisclaw.ai/v1/marketplace/compute/sandbox/exec \
  -H "Authorization: Bearer sk-your-api-key" \
  -H "Content-Type: application/json" \
  -d '{"sandbox_id": "sb_a1b2c3d4e5", "command": "python -c \"print(2+2)\""}'

# Check status
curl -X POST https://api.jarvisclaw.ai/v1/marketplace/compute/sandbox/status \
  -H "Authorization: Bearer sk-your-api-key" \
  -H "Content-Type: application/json" \
  -d '{"sandbox_id": "sb_a1b2c3d4e5"}'

# Terminate
curl -X POST https://api.jarvisclaw.ai/v1/marketplace/compute/sandbox/terminate \
  -H "Authorization: Bearer sk-your-api-key" \
  -H "Content-Type: application/json" \
  -d '{"sandbox_id": "sb_a1b2c3d4e5"}'

python

import requests

BASE = "https://api.jarvisclaw.ai/v1/marketplace/compute/sandbox"
HEADERS = {
    "Authorization": "Bearer sk-your-api-key",
    "Content-Type": "application/json",
}

# Create a sandbox
resp = requests.post(f"{BASE}/create", headers=HEADERS, json={
    "timeout": 300,
})
sandbox_id = resp.json()["sandbox_id"]
print(f"Created sandbox: {sandbox_id}")

# Execute a command
resp = requests.post(f"{BASE}/exec", headers=HEADERS, json={
    "sandbox_id": sandbox_id,
    "command": "python -c \"import math; print(math.pi)\"",
})
print(resp.json()["stdout"])

# Check status
resp = requests.post(f"{BASE}/status", headers=HEADERS, json={
    "sandbox_id": sandbox_id,
})
print(resp.json()["status"])

# Terminate when done
requests.post(f"{BASE}/terminate", headers=HEADERS, json={
    "sandbox_id": sandbox_id,
})

python

from jarvisclaw import ComputeClient

# --- Option A: Base chain (EVM) ---
# Hex private key -> USDC on Base (Chain ID 8453)
compute = ComputeClient(private_key="0x<evm-private-key>")

# --- Option B: Solana ---
# Base58 keypair -> USDC SPL on Solana mainnet
# compute = ComputeClient(private_key="<solana-bs58-keypair>")

# SDK auto-detects chain from key format - no config needed

# Create a sandbox
sandbox = compute.create(timeout=300)
print(f"Sandbox ID: {sandbox.sandbox_id}")

# Execute commands
result = compute.exec(sandbox.sandbox_id, "pip install numpy && python -c \"import numpy; print(numpy.__version__)\"")
print(result.stdout)

result = compute.exec(sandbox.sandbox_id, "python -c \"import numpy as np; print(np.random.rand(3))\"")
print(result.stdout)

# Check status
status = compute.status(sandbox.sandbox_id)
print(f"Commands run: {status.commands_executed}")

# Terminate
compute.terminate(sandbox.sandbox_id)

package main

import (
    "context"
    "fmt"
    jc "github.com/api-jarvisclaw/go-sdk"
)

func main() {
    ctx := context.Background()
    cc, _ := jc.NewComputeClient(jc.WithAPIKey("sk-your-api-key"))

    // Create a sandbox
    sandbox, _ := cc.Create(ctx, jc.WithTimeout(300))
    fmt.Printf("Sandbox ID: %s\n", sandbox.SandboxID)

    // Execute a command
    result, _ := cc.Exec(ctx, sandbox.SandboxID, "python -c \"print('hello')\"")
    fmt.Printf("Output: %s\n", result.Stdout)

    // Check status
    status, _ := cc.Status(ctx, sandbox.SandboxID)
    fmt.Printf("Status: %s, Commands: %d\n", status.Status, status.CommandsExecuted)

    // Terminate
    cc.Terminate(ctx, sandbox.SandboxID)
}

package main

import (
    "context"
    "fmt"
    jc "github.com/api-jarvisclaw/go-sdk"
)

func main() {
    ctx := context.Background()

    // x402 Agent wallet - pays per-call via USDC on Base (Chain ID 8453)
    cc, _ := jc.NewComputeClient(jc.WithPrivateKey("0x<evm-private-key>"))

    // Create a sandbox
    sandbox, _ := cc.Create(ctx, jc.WithTimeout(300))
    fmt.Printf("Sandbox ID: %s\n", sandbox.SandboxID)

    // Execute commands
    result, _ := cc.Exec(ctx, sandbox.SandboxID, "python -c \"print('agent execution')\"")
    fmt.Printf("Output: %s\n", result.Stdout)

    // Terminate when done
    cc.Terminate(ctx, sandbox.SandboxID)
}

Limitations

24h max timeout — sandboxes auto-terminate after the configured timeout (max 86400s)
Flat-rate for short tasks — ≤300s gets flat CPU $0.01, longer sessions billed per hour
No persistent storage — all data is lost when the sandbox terminates
No inbound connections — sandboxes cannot receive network traffic from outside
10MB max payload — request and response bodies are capped at 10MB
Outbound HTTP allowed — sandboxes can make outbound requests (pip install, curl, etc.)

Compute API (Modal Sandbox) ​

Endpoints ​

Pricing ​

Burst — timeout ≤ 300s ​

Long-lived — timeout > 300s, up to 24h ​

POST /sandbox/create ​

Parameters ​

Request ​

Response ​

POST /sandbox/exec ​

Parameters ​

Request ​

Response ​

POST /sandbox/status ​

Parameters ​

Request ​

Response ​

POST /sandbox/terminate ​

Parameters ​

Request ​

Response ​

Errors ​

Code Examples ​

Limitations ​

Compute API (Modal Sandbox)

Endpoints

Pricing

Burst — timeout ≤ 300s

Long-lived — timeout > 300s, up to 24h

POST /sandbox/create

Parameters

Request

Response

POST /sandbox/exec

Parameters

Request

Response

POST /sandbox/status

Parameters

Request

Response

POST /sandbox/terminate

Parameters

Request

Response

Errors

Code Examples

Limitations