unsandbox.com
Anonymous remote code, compile, & execution API for humans & machine learning agents.
Docs 📚 View Pricing →The Julia examples below require the following 3rd party libraries:
HTTP.jlJSON.jl
Make sure these packages are installed in your environment before running the examples.
Execute code immediately and wait for results. Best for quick scripts and interactive use.
Request Parameters
| Parameter | Type | Required | Description |
|---|---|---|---|
language |
string | ✓ | Programming language |
code |
string | ✓ | Source code to execute |
env |
object | Environment variables (key-value pairs) | |
network_mode |
string | "zerotrust" or "semitrusted" | |
ttl |
integer | Timeout 1-900s (default: 60) | |
return_artifact |
boolean | Return compiled binary | |
return_wasm_artifact |
boolean | Return WebAssembly binary | |
input_files |
array | Files to upload (see Input Files) |
Example (Julia)
using HTTP
using JSON
api_key = get(ENV, "UNSANDBOX_API_KEY", "unsb-sk-test0-vault-unlck-12345")
url = "https://api.unsandbox.com/execute"
headers = Dict(
"Content-Type" => "application/json",
"Authorization" => "Bearer $api_key"
)
payload = JSON.json(Dict(
"language" => "julia",
"code" => "println(\"Hello from unsandbox!\")"
))
response = HTTP.post(url, headers, payload)
result = JSON.parse(String(response.body))
if result["success"]
println("Output: ", result["stdout"])
else
println("Error: ", result["error"])
end
Response (Success)
{
"success": true,
"stdout": "Hello from unsandbox!\n",
"stderr": "",
"exit_code": 0,
"language": "python",
"job_id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
"artifacts": [],
"execution_mode": "pooled",
"network_mode": "zerotrust",
"total_time_ms": 523
}
Response (Error)
{
"success": false,
"stdout": "",
"stderr": "SyntaxError: invalid syntax\n",
"exit_code": 1,
"language": "python",
"job_id": "b2c3d4e5-f6a7-8901-bcde-f23456789012",
"artifacts": [],
"execution_mode": "pooled",
"network_mode": "zerotrust",
"total_time_ms": 312
}
Submit code for execution and receive a job ID to poll for results. Best for long-running scripts or when you need to decouple submission from execution.
Request Parameters
| Parameter | Type | Required | Description |
|---|---|---|---|
language |
string | ✓ | Programming language |
code |
string | ✓ | Source code to execute |
env |
object | Environment variables (key-value pairs) | |
network_mode |
string | "zerotrust" or "semitrusted" | |
ttl |
integer | Timeout 1-900s (default: 60) | |
return_artifact |
boolean | Return compiled binary | |
return_wasm_artifact |
boolean | Return WebAssembly binary | |
input_files |
array | Files to upload (see Input Files) |
Example (Julia)
using HTTP
using JSON
api_key = get(ENV, "UNSANDBOX_API_KEY", "unsb-sk-test0-vault-unlck-12345")
headers = Dict(
"Content-Type" => "application/json",
"Authorization" => "Bearer $api_key"
)
# Submit job
url = "https://api.unsandbox.com/execute/async"
payload = JSON.json(Dict(
"language" => "rust",
"code" => "fn main() { println!(\"Computing...\"); }",
"ttl" => 300
))
response = HTTP.post(url, headers, payload)
job = JSON.parse(String(response.body))
job_id = job["job_id"]
println("Job submitted: ", job_id)
# Wait briefly for job to be created
sleep(2)
# Poll for results
while true
status_url = "https://api.unsandbox.com/jobs/$job_id"
status_response = HTTP.get(status_url, headers)
status = JSON.parse(String(status_response.body))
if status["status"] == "completed"
result = get(status, "result", nothing)
if result !== nothing && get(result, "success", false)
println("Output: ", result["stdout"])
elseif result !== nothing
println("Error: ", result["stderr"])
else
println("Error: No result available")
end
break
elseif status["status"] in ["timeout", "cancelled"]
println("Job ", status["status"])
break
end
sleep(1)
end
Initial Response
{
"job_id": "job_1234567890_abc",
"status": "pending"
}
Check the status and results of an asynchronous job. Poll this endpoint after submitting a job via /execute/async or /run/async.
URL Parameters
| Parameter | Type | Description |
|---|---|---|
id |
string | Job ID returned from async endpoint |
Example (Julia)
using HTTP
using JSON
api_key = get(ENV, "UNSANDBOX_API_KEY", "unsb-sk-test0-vault-unlck-12345")
headers = Dict(
"Content-Type" => "application/json",
"Authorization" => "Bearer $api_key"
)
# Submit a job to get a real job ID
submit_url = "https://api.unsandbox.com/execute/async"
payload = JSON.json(Dict(
"language" => "python",
"code" => "print('Example job')"
))
submit_response = HTTP.post(submit_url, headers, payload)
job_data = JSON.parse(String(submit_response.body))
job_id = job_data["job_id"]
println("Submitted job: ", job_id)
# Get job status
status_url = "https://api.unsandbox.com/jobs/$job_id"
status_headers = Dict("Authorization" => "Bearer $api_key")
response = HTTP.get(status_url, status_headers)
job = JSON.parse(String(response.body))
println("Status: ", job["status"])
if job["status"] == "completed"
println("Exit code: ", get(job, "exit_code", "N/A"))
println("Output: ", get(job, "stdout", ""))
end
Response (Completed)
{
"job_id": "c3d4e5f6-a7b8-9012-cdef-345678901234",
"status": "completed",
"created_at": 1699564800,
"started_at": 1699564801,
"completed_at": 1699564805,
"result": {
"success": true,
"stdout": "Hello from unsandbox!\n",
"stderr": "",
"exit_code": 0,
"language": "python",
"artifacts": [],
"execution_mode": "pooled",
"network_mode": "zerotrust",
"total_time_ms": 4021
}
}
Possible Status Values
pending- Job queued, waiting to executerunning- Job currently executingcompleted- Job finished (check result field)timeout- Job exceeded TTL limitcancelled- Job was cancelled via DELETE
List all active (pending or running) jobs for your API key. Useful for monitoring multiple async executions.
Example (Julia)
using HTTP
using JSON
api_key = get(ENV, "UNSANDBOX_API_KEY", "unsb-sk-test0-vault-unlck-12345")
url = "https://api.unsandbox.com/jobs"
headers = Dict("Authorization" => "Bearer $api_key")
response = HTTP.get(url, headers)
jobs = JSON.parse(String(response.body))
for job in jobs
println("$(job["job_id"]): $(job["status"]) ($(job["language"]))")
end
Response
{
"jobs": [
{
"job_id": "job_1234567890_abc",
"status": "running",
"language": "python",
"submitted_at": "2024-01-15T10:30:00Z"
},
{
"job_id": "job_0987654321_xyz",
"status": "pending",
"language": "go",
"submitted_at": "2024-01-15T10:29:45Z"
}
]
}
Cancel a pending or running job. If the job is already executing, it will be terminated and partial output returned.
URL Parameters
| Parameter | Type | Description |
|---|---|---|
id |
string | Job ID to cancel |
Example (Julia)
using HTTP
using JSON
api_key = get(ENV, "UNSANDBOX_API_KEY", "unsb-sk-test0-vault-unlck-12345")
headers = Dict(
"Content-Type" => "application/json",
"Authorization" => "Bearer $api_key"
)
# Submit a long-running job to demonstrate cancellation
submit_url = "https://api.unsandbox.com/execute/async"
payload = JSON.json(Dict(
"language" => "python",
"code" => "import time; time.sleep(10); print('Done')"
))
submit_response = HTTP.post(submit_url, headers, payload)
job_data = JSON.parse(String(submit_response.body))
job_id = job_data["job_id"]
println("Submitted job: ", job_id)
# Delete/cancel the job
delete_url = "https://api.unsandbox.com/jobs/$job_id"
delete_headers = Dict("Authorization" => "Bearer $api_key")
response = HTTP.delete(delete_url, delete_headers)
result = JSON.parse(String(response.body))
println("Cancelled: ", get(result, "message", "Job cancelled"))
if haskey(result, "stdout")
println("Partial output: ", result["stdout"])
end
Response
{
"message": "Job cancelled",
"job_id": "e5f6a7b8-c9d0-1234-efgh-567890123456",
"result": {
"success": false,
"stdout": "Partial output before cancel...",
"stderr": "",
"exit_code": -1,
"language": "python",
"artifacts": [],
"execution_mode": "pooled",
"network_mode": "zerotrust",
"total_time_ms": 1523
}
}
Execute code with automatic language detection via shebang. Send raw code as the request body with Content-Type: text/plain.
Request
- Content-Type:
text/plain - Body: Raw code with shebang (e.g.,
#!/usr/bin/env python3)
Query Parameters
| Parameter | Type | Required | Description |
|---|---|---|---|
env |
string | URL-encoded JSON object with environment variables | |
network_mode |
string | "zerotrust" or "semitrusted" | |
ttl |
integer | Timeout 1-900s (default: 60) | |
return_artifact |
boolean | Return compiled binary | |
return_wasm_artifact |
boolean | Return WebAssembly binary |
Example (Julia)
using HTTP
using JSON
api_key = get(ENV, "UNSANDBOX_API_KEY", "unsb-sk-test0-vault-unlck-12345")
url = "https://api.unsandbox.com/run"
headers = Dict(
"Content-Type" => "text/plain",
"Authorization" => "Bearer $api_key"
)
code = """#!/usr/bin/env python3
print("Language auto-detected!")"""
response = HTTP.post(url, headers, code)
result = JSON.parse(String(response.body))
println("Detected: ", result["detected_language"])
println("Output: ", result["stdout"])
Response
{
"success": true,
"stdout": "Language auto-detected!\n",
"stderr": "",
"exit_code": 0,
"language": "python",
"detected_language": "python",
"job_id": "f6a7b8c9-d0e1-2345-fghi-678901234567",
"artifacts": [],
"execution_mode": "pooled",
"network_mode": "zerotrust",
"total_time_ms": 412
}
Submit code with automatic language detection and receive a job ID. Combines the convenience of /run with the flexibility of async execution.
Request
- Content-Type:
text/plain - Body: Raw code with shebang (e.g.,
#!/usr/bin/env ruby)
Query Parameters
| Parameter | Type | Required | Description |
|---|---|---|---|
env |
string | URL-encoded JSON object with environment variables | |
network_mode |
string | "zerotrust" or "semitrusted" | |
ttl |
integer | Timeout 1-900s (default: 60) | |
return_artifact |
boolean | Return compiled binary | |
return_wasm_artifact |
boolean | Return WebAssembly binary |
Example (Julia)
using HTTP
using JSON
api_key = get(ENV, "UNSANDBOX_API_KEY", "unsb-sk-test0-vault-unlck-12345")
headers = Dict(
"Content-Type" => "text/plain",
"Authorization" => "Bearer $api_key"
)
code = """#!/usr/bin/env ruby
puts "Running async with auto-detect!" """
# Submit job
url = "https://api.unsandbox.com/run/async"
response = HTTP.post(url, headers, code)
job = JSON.parse(String(response.body))
job_id = job["job_id"]
println("Job submitted: ", job_id)
# Wait briefly for job to be created
sleep(2)
# Poll for results
while true
status_url = "https://api.unsandbox.com/jobs/$job_id"
status_response = HTTP.get(status_url, headers)
status = JSON.parse(String(status_response.body))
if status["status"] == "completed"
result = get(status, "result", nothing)
if result !== nothing
println("Detected: ", result["detected_language"])
if get(result, "success", false)
println("Output: ", result["stdout"])
end
else
println("Error: No result available")
end
break
elseif status["status"] in ["timeout", "cancelled"]
println("Job ", status["status"])
break
end
sleep(1)
end
Initial Response
{
"job_id": "job_1234567890_abc",
"status": "pending"
}
Job Result (via GET /jobs/{id})
{
"job_id": "a7b8c9d0-e1f2-3456-ghij-789012345678",
"status": "completed",
"created_at": 1699564800,
"started_at": 1699564801,
"completed_at": 1699564802,
"result": {
"success": true,
"stdout": "Running async with auto-detect!\n",
"stderr": "",
"exit_code": 0,
"language": "ruby",
"detected_language": "ruby",
"artifacts": [],
"execution_mode": "pooled",
"network_mode": "zerotrust",
"total_time_ms": 1234
}
}
Upload files to your sandbox execution. Files are written to the working directory and can be accessed directly by filename.
File Object Schema
| Field | Type | Required | Description |
|---|---|---|---|
filename |
string | Filename (e.g., "data.csv"). If omitted, file is named by index: "0", "1", etc. | |
content |
string | ✓ | Base64-encoded file content |
Example Request
{
"language": "python",
"code": "print(open('data.csv').read())",
"input_files": [
{
"filename": "data.csv",
"content": "bmFtZSxhZ2UKQWxpY2UsMzAKQm9iLDI1"
}
]
}
Limits
- Maximum files: 10 files per execution
- Maximum total size: 10MB (decoded)
- Maximum single file: 5MB (decoded)
- Filename length: 256 characters max
File Access
Files are placed in the working directory. Access them directly:
- Python:
open('data.csv') - Node.js:
fs.readFileSync('data.csv') - Ruby:
File.read('data.csv') - Go:
os.ReadFile("data.csv")
Security
- Files are isolated per execution
- Filenames cannot contain path separators (no
/or..) - Files are deleted after execution completes