At a glance
| Version | Released | Highlights |
|---|
| 3.4.0 | Phase 2 | Direct API channel, Stripe self-serve, Mega tier (200K exec/mo) |
| 3.3.0 | Phase 2 | Package installation in sessions (pip / npm / gem / cargo) with offline cache |
| 3.2.0 | Phase 2 | Stateful sessions — REPL-like execution with persistent state |
| 3.1.0 | Phase 2 | SSE streaming output; async + signed webhook callbacks |
| 3.0.0 | Phase 2 | 4 new languages, multi-file programs, expected_output, MCP expanded to 11 tools |
| 2.0.0 | Phase 1 | Ultra tier, gVisor isolation, batch up to 100, 300s timeout, playground |
| 1.2.0 | Phase 1 | Pre-warmed sandbox pools, sub-100ms cold starts, memory reporting |
| 1.1.0 | Phase 1 | Batch execution endpoint |
| 1.0.0 | Phase 1 | Stable release — Pro tier, rate-limit headers, /health, full error codes |
| 0.3.0 | Phase 1 | All 8 languages, stdin support, /languages endpoint |
| 0.2.0 | Phase 1 | Added TypeScript, Go, Bash |
| 0.1.0 | Phase 1 | Alpha — Python and JavaScript |
Direct API channel + Mega tier
SandboxAPI is now available outside RapidAPI for teams that want lower latency and self-serve billing. The new api.sandboxapi.dev channel uses Stripe for billing and supports programmatic key rotation and webhook IP allowlisting. The Mega tier ships at $149/mo with 200,000 executions, 600s timeouts, batches up to 200, and 50 concurrent sessions — designed for institutional assessment platforms and large agent fleets.
- Direct API at
api.sandboxapi.dev bypasses RapidAPI for ~30ms lower latency.
- Stripe-backed self-serve billing on the direct channel.
- New Mega tier (200K exec/mo, 600s, batch 200, 50 sessions, custom allowlist).
- Programmatic API-key rotation endpoint for direct customers.
Package installation in sessions
The piece that makes code interpreters credible. Inside any session, your agent can pip install / npm install / gem install / cargo add packages. Top-1k packages per language are cached for sub-3-second installs. Network egress for the install step is gated by a per-request allowlist proxy.
- New
POST /v1/sessions/{id}/install endpoint.
- Supported managers:
pip, npm, gem, cargo.
- Top-1k packages cached locally for fast installs.
- MCP
session_install_packages tool.
Stateful sessions
Persistent sandboxes that retain variables, files, and installed packages across calls. The single biggest unlock for AI-agent code interpreters: variables you set in one call are still there in the next. Sessions auto-renew on activity and expire at the idle TTL.
POST /v1/sessions, POST /v1/sessions/{id}/execute, GET /v1/sessions/{id}, DELETE /v1/sessions/{id}.- Per-API-key concurrent session limits enforced by tier (5 / 20 / 50).
- Idle TTL up to 30 minutes; auto-renew on execute and install.
- MCP tools:
create_session, session_execute, session_close.
Streaming output and async + webhooks
Two long-requested integration modes shipped together. Streaming uses Server-Sent Events to deliver live stdout/stderr as code runs. Async lets you fire-and-forget long jobs and get a signed callback when they complete. Webhook payloads are HMAC-SHA256 signed.
POST /v1/execute/stream — SSE events: stdout, stderr, result.POST /v1/execute?async=true — returns a token; results polled at /v1/executions/{token} or pushed to callback_url.- Signed webhooks via
X-SandboxAPI-Signature header.
- MCP
execute_async and get_execution tools.
4 new languages, multi-file, output verification
The "we caught up to Judge0" release. Adds Rust 1.85, Ruby 3.3, PHP 8.4, and C# (.NET 9), bringing the total to 12 modern languages. Multi-file submissions accept a base64-encoded ZIP plus optional compile/run scripts. The new expected_output field returns a wrong_answer status when stdout doesn't match — the canonical online-judge primitive. The MCP server expands from 5 tools to 11.
- Languages added: Rust 1.85, Ruby 3.3, PHP 8.4, C# (.NET 9).
- Multi-file:
additional_files (base64 ZIP), compile_script, run_script.
expected_output field with wrong_answer status.- Status taxonomy:
completed, wrong_answer, timeout, compilation_error, runtime_error, memory_limit, internal_error.
- MCP server now exposes 11 tools (was 5).
- New fields:
compiler_options, command_line_arguments, redirect_stderr_to_stdout, base64_encoded.
Ultra tier, gVisor, and the playground
Phase 1's flagship release. Adds the Ultra tier (50K exec/mo), enables gVisor isolation as the default runtime, raises batch size to 100, lifts the timeout to 300s, and ships the public playground at sandboxapi.dev/playground.
Sandbox pre-warming
Pre-warmed sandbox pools eliminate cold-start latency. p50 cold-start drops below 100ms across all 8 supported languages. Adds memory_used_kb to the response.
Batch execution
New POST /v1/execute/batch endpoint runs up to 50 snippets in a single round trip — perfect for grading and bulk validation.
Stable release
API surface frozen. Pro tier launches. Standard rate-limit headers (X-RateLimit-Remaining, X-RateLimit-Reset) included on every response. GET /health endpoint and the full error-code taxonomy land.
All 8 launch languages, stdin
Adds Java, C, C++ to bring Phase 1 language coverage to 8. Adds stdin support and the /v1/languages discovery endpoint.
TypeScript, Go, Bash
First language expansion. TypeScript transpiles via tsc, Go compiles per-call, Bash runs via the system shell — all under gVisor.
Alpha — Python and JavaScript
The first cut. Synchronous execution only, two languages, no batch, no stdin. Just enough to prove the model.