Together Code Sandbox vs Agent 37

Together Code Sandbox starts VMs from Docker-image snapshots on Together's AI cloud, and Together must turn it on for your organization. Agent 37 is self-serve hosting for always-on agents, one gVisor sandbox per user, that sleep when idle and wake on the next message.

The companies and products

Together Code Sandbox

Together AI is a cloud for running and training AI models, and Code Sandbox is its product for running code in isolated VMs.

You build a snapshot from a Docker image or Dockerfile on Together's builder, then start sandboxes from it and run commands and file operations through the together-sandbox SDK or CLI. When a sandbox ends you can save its disk, and its memory too, as a new snapshot to start more sandboxes from. It is aimed at model-generated code, agent tool calls and reinforcement learning rollouts, and it must be enabled for your organization by Together. Together also sells Code Interpreter, a separate API that runs Python code in short sessions.

Company: Together AI, a cloud for model inference, fine-tuning and GPU clusters
Isolation: A virtual machine per sandbox on Together's infrastructure
Access: The SDK and CLI must be enabled for your organization; you ask Together for access
Machine size: 0.1 to 16 cores, 1 to 8 GB of memory per core; disk size is not configurable
History: Built on CodeSandbox, now a Together company; the CodeSandbox-based SDK is deprecated and together-sandbox is its successor
Source code: The SDKs and CLI are MIT-licensed on GitHub; the platform itself is not

Agent 37

Agent 37 hosts AI agents, one isolated, always-on sandbox per user, behind one API.

You create an instance from a template (Hermes, OpenClaw, Claude Code, Codex, OpenCode, Pi, Grok or your own Docker image) and talk to it through a chat API, files, exec, SSH and HTTPS URLs. The default instance is a container enforced by gVisor; a VM type gives a full Linux machine with its own kernel for Docker, Kubernetes and Android emulators. It is built for companies that give each of their customers an agent.

Isolation: gVisor by default; opt-in full VM type with its own kernel and KVM
Where it runs: Dedicated servers in the United States
Billing: Prepaid wallet, metered per minute, no plan fee or seat charge
Instance types: Default (shared), Performance (dedicated cores), VM
Compliance: SOC 2 Type I, Type II in progress
Company: Y Combinator company based in New York

Which use cases each one fits

For each use case, Battleships prices every sandbox provider on the same workload and sorts them from the cheapest monthly bill up. “65th cheapest of 367” means 64 of the other 366 providers run that workload, with every feature it needs, for less. “With limits” means the provider runs it but misses something, so it is listed after every full match. Open a use case for its whole ranking and what each provider misses.

Use case and its workload	Together Code Sandbox: price rank	Agent 37: price rank	bochka: price rank
Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month Neither gives Docker on its default setup: Agent 37 needs its VM type at 2.5x compute, and Together does not document Docker inside; bochka has Docker, systemd and /dev/kvm on every machine.	65th of 367, with limits $528 a month	2nd cheapest of 367 $551 a month	3rd cheapest of 367 $634 a month
Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB Agent 37 is built for one always-on assistant per user that sleeps, keeps its memory and wakes on a message; Together sandboxes have no idle detection or wake on request.	152nd of 366, with limits $1,095 a month	Cheapest of 366 $279 a month	5th cheapest of 366 $657 a month
Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month Together's Code Interpreter runs short Python sessions billed per 60 minutes; Agent 37's small default sandboxes run Python or Node and keep their files.	326th of 345, with limits $2,190 a month	Cheapest of 345 $22 a month	186th of 345, with limits $2,000 a month
Sandboxes for your users 300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month Agent 37 stops at 200 instances on its top self-serve tier, sleeping ones included; Together makes every port public; bochka gives token-protected HTTPS URLs and plans up to 2,000 at once.	234th of 361, with limits $900 a month	30th of 361, with limits $79 a month	Cheapest of 361 $209 a month
RL rollouts 1,000 at once · 2 vCPU / 4 GB · 1,000,000 sessions of 3 min a month Together can save a sandbox with its memory and start many copies from it, which suits rollouts that branch from one warm state; Agent 37 has no fork, and bochka forks a running sandbox.	315th of 360, with limits $21,900 a month	6th of 360, with limits $363 a month	34th of 360, with limits $2,000 a month
Agent evals 250 at once · 4 vCPU / 8 GB · 15,000 sessions of 20 min a month	258th of 364, with limits $450 a month	90th of 364, with limits $58 a month	10th cheapest of 364 $200 a month
CI runners 40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month	251st of 377, with limits $180 a month	2nd cheapest of 377 $84 a month	4th cheapest of 377 $86 a month
24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB Agent 37 keeps a machine and its disk around the clock and is cheap to keep idle; Together's self-serve product keeps nothing after a session.	210th of 368, with limits $22 a month	3rd cheapest of 368 $12 a month	18th cheapest of 368 $26 a month
Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month Agent 37 adds a noVNC desktop through a template build and Together has no desktop; bochka streams a 60fps desktop with Chromium in every sandbox.	184th of 369, with limits $600 a month	Cheapest of 369 $53 a month	6th cheapest of 369 $200 a month
Browser-use agents 50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month	301st of 367, with limits $900 a month	48th of 367, with limits $29 a month	7th cheapest of 367 $200 a month
Scraping & bots (startup scale) 50 at once · 1 vCPU / 2 GB · 100,000 sessions of 3 min a month	219th of 346, with limits $180 a month	Cheapest of 346 $32 a month	6th cheapest of 346 $90 a month
Mass scraping & bots (undetectable) 1,000 at once · 1 vCPU / 2 GB · 3,000,000 sessions of 2 min a month	124th of 338, with limits $3,600 a month	36th of 338, with limits $648 a month	71st of 338, with limits $1,800 a month
Parallel red teaming 200 at once · 2 vCPU / 4 GB · 5,000 sessions of 30 min a month Agent 37 runs each attempt in its own sandbox but cannot fork one; bochka can fork a running VM to try many paths at once.	167th of 360, with limits $150 a month	Cheapest of 360 $43 a month	2nd cheapest of 360 $200 a month
GPU eval jobs 10 at once · 8 vCPU / 32 GB · 500 sessions of 30 min a month	Can't run it	Can't run it	Can't run it
macOS agents 5 at once · 4 vCPU / 8 GB · 1,000 sessions of 20 min a month	Can't run it	Can't run it	Can't run it

Key features

Feature	Together Code Sandbox	Agent 37	bochka
Getting started	The SDK and CLI must be enabled for your organization by Together	Sign up and top up a prepaid wallet; one instance before the first top-up	Sign up and start a sandbox from the API, CLI or dashboard
Isolation	A VM per sandbox on Together's infrastructure	gVisor by default; full VM as an opt-in instance type	Every sandbox is a full VM with its own kernel on dedicated AMD Ryzen 9 9950X hosts
Machine size	0.1 to 16 cores, 1 to 8 GB of memory per core; disk size fixed	Fixed shapes from 2 vCPU / 4 GB up to 16 vCPU / 32 GB; a 4 vCPU / 8 GB instance takes at most 20 GB of disk	2 to 16 vCPU with 2 GB per vCPU, up to 16 vCPU / 32 GB on the $100 plan and above
Docker inside	Not documented; Dev Containers and Docker Compose are not read	Only on the VM type, at 2.5x the default compute rate	Yes, in every sandbox
Root and systemd	Commands run as user 1000 by default, root on request	Root over SSH; your entrypoint is PID 1 with no init system	Root and systemd on every machine
Nested virtualization	Not documented	/dev/kvm on the VM type, with a ready Android template	/dev/kvm in every sandbox, Android emulator works
Shell access	Commands, interactive shells and files through the SDK and CLI; no SSH	SSH through a WebSocket tunnel, plus exec and files over the API	SSH, plus exec and files over the API
Saving memory	End a sandbox with a snapshot of its disk and memory; new sandboxes resume with processes running	Memory checkpointed on auto-sleep, processes resume on wake	Disk saved every minute and on stop; a stopped sandbox costs nothing; resume reboots the machine
Fork and copies	No live fork: the parent stops, then you start many sandboxes from its snapshot	No fork; daily backups plus one on-demand backup, restored in place only	Fork a running sandbox while the original keeps running; named snapshots as shareable templates
Idle and wake	Time to live counts from creation; no idle detection and no wake on request	Opt-in auto-sleep after a quiet period (default 15 minutes), wakes on the next request	No idle timer; a time limit you set, or no limit; started again by API
Reaching the sandbox	All ports public; preview URLs coming soon; an experimental inbound allowlist	HTTPS URL per port, signed URLs, custom domains; no public IP	HTTPS URLs for up to 50 ports, token-protected by default; raw TCP; dedicated IPv6 or IPv4
Desktop and browser	None; headless command and file access only	noVNC desktop through a template build or the Ubuntu VM template	60fps desktop and Chromium in every sandbox, embeddable stream
Coding agents	Run your own agent through execs; a Harbor guide for agent evals	Hermes, OpenClaw, Claude Code, Codex, Grok, OpenCode, Pi, with a chat API per instance	Claude Code, Codex, pi, OpenCode and more preinstalled, with a prompt, steer and events API
Keeping API keys out of the sandbox	No credentials proxy documented	Managed models, search and Composio through a per-instance token; your own keys go in as env vars	Proxy for model providers and any HTTPS API; switch a credential to another account when credits run out

Every feature Battleships tracks (43)

Feature	Together Code Sandbox	Agent 37	bochka
Full VM (own kernel)	Yes	No	Yes
Docker inside	Not published	No	Yes
Nested virtualization	No	No	Yes
Root	Yes	Yes	Yes
Systemd	Not published	Not published	Yes
Desktop GUI	No	Yes	Yes
Browser	No	Yes	Yes
Computer-use API	No	Yes	Yes
SSH	No	Yes	Yes
Persistent disk	Yes	Yes	Yes
Pause/resume	Yes	Yes	Yes
Automatic snapshots	No	Yes	Yes
Fork/clone	Yes	Yes	Yes
Live fork (no pause)	No	Yes	Yes
Memory fork	Yes	No	No
Custom image (Docker or snapshot)	Yes	Yes	Yes
≥24 h sessions	Yes	Yes	Yes
Idle auto-stop	No	Yes	No
Wake on request	No	Yes	No
HTTPS preview URLs	Yes	Yes	Yes
Raw TCP inbound	No	Not published	Yes
Public IPv4	No	No	No
Egress allowlist	No	No	No
Secret proxy	No	Yes	Yes
Secret proxy for any API	No	No	Yes
Preinstalled agents	No	Yes	Yes
Agent harness API	No	Yes	Yes
MCP server	Yes	No	No
GPU	No	No	No
Windows	No	No	No
Arm64	No	No	No
Self-hosting	No	No	No
BYOC	No	No	No
Open source	No	No	No
Scoped API keys	Not published	Not published	Yes
Spend limits	Yes	Yes	Yes
Usage per sandbox (API)	Yes	Yes	Yes
Audit logs	Not published	Not published	No
SSO	Yes	No	No
SOC 2	Yes	Yes	No
HIPAA	No	No	No
EU data residency	Not published	No	Yes
Zero data retention	Not published	Not published	Yes

Performance

Together Code SandboxNot measured

Agent 37Not measured

bochka26.69 runs/s

Neither Together Code Sandbox nor Agent 37 has a comparable CPU measurement. bochka did 26.69 runs per second on the StarSling hpc-sandbox-benchmarks Node.js test at 4 vCPU / 8 GB, on dedicated AMD Ryzen 9 9950X hosts. Agent 37 says its default instances share host resources and sells a Performance type on dedicated cores.

Node.js build-and-test loop at 4 vCPU / 8 GB, higher is better. Measured with the open hpc-sandbox-benchmarks suite. Together Code Sandbox and Agent 37 have no published run in this suite yet, so there is no like-for-like number to compare.

Pricing in five scenarios

Each bill is priced from the providers' public price lists by the open Battleships engine, with the cheapest plan that runs the scenario.

Scenario	Together Code Sandbox, a month	Agent 37, a month	bochka, a month
One coding agent, workdays 4 vCPU / 8 GB, 50 GB disk, 8 hours a day, 22 days a month, files kept between days	$5 · Together pay-as-you-go (prepaid credits; Code Interpreter only) · with limits	$6 · Wallet (any top-up)	$20 · $20/mo
One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month	$22 · Together pay-as-you-go (prepaid credits; Code Interpreter only) · with limits	$12 · Wallet (any top-up)	$26 · $20/mo
50 agents in parallel 50 machines of 4 vCPU / 8 GB at once, 8-hour sessions, 1,100 sessions a month, Docker inside	$264 · Together pay-as-you-go (prepaid credits; Code Interpreter only) · with limits	$329 · Wallet ($100 lifetime top-ups)	$317 · $20/mo × 3 seats
Sandboxes for your users 2 vCPU / 4 GB, 30,000 sessions of 10 minutes a month, up to 300 at once	$900 · Together pay-as-you-go (prepaid credits; Code Interpreter only) · with limits	$33 · Wallet ($250 lifetime top-ups) · with limits	$200 · $100/mo × 2 seats
1,000 short jobs at once 2 vCPU / 4 GB, 1,000 at once, 3 minutes each, 200,000 jobs a month	$6,000 · Together pay-as-you-go (prepaid credits; Code Interpreter only) · with limits	$73 · Wallet ($250 lifetime top-ups) · with limits	$2,000 · $2000/mo · with limits

Together Code Sandbox: Together lists Code Sandbox at $0.0446 per vCPU-hour and $0.0149 per GiB-hour, but the VMs must be enabled for your organization, so the table prices Code Interpreter: $0.03 per 60-minute session from prepaid credits, with a $5 minimum.

Agent 37: A prepaid wallet with no plan fee, metered per minute: $0.80 per vCPU and $0.70 per GB of RAM per month while running, and $0.09 per GB of disk per month even while asleep. VM compute costs 2.5x and Performance 4x the default rate.

Agent 37: Plans start at $20 a month, and the whole fee becomes usage credit. You pay per second while a sandbox runs, and a stopped sandbox costs nothing and keeps its disk.

Limits in those scenarios

The Together column is Code Interpreter, and it fits none of the scenarios: sessions end after an hour, nothing is kept, it runs only Python, and the machine size is not published. Agent 37 fits the three single-team scenarios, but its cap of 200 instances, counting sleeping and stopped ones, rules it out for 300 user sandboxes at once and for 1,000 short jobs. The 50-agent row needs Docker, so Agent 37 is priced on its VM type there. bochka's limit is starts per hour and per day, which stops it in the 1,000-at-once burst.

Scenario	Together Code Sandbox	Agent 37	bochka
One coding agent, workdays	Together pay-as-you-go (prepaid credits; Code Interpreter only): sessions capped at 1 h missing persistent disk machine size not published different kind of product: code interpreter	egress price not published	None that bite
One agent that never stops	Together pay-as-you-go (prepaid credits; Code Interpreter only): sessions capped at 1 h missing persistent disk machine size not published different kind of product: code interpreter	egress price not published	None that bite
50 agents in parallel	Together pay-as-you-go (prepaid credits; Code Interpreter only): sessions capped at 1 h missing persistent disk, Docker inside machine size not published different kind of product: code interpreter	egress price not published	None that bite
Sandboxes for your users	machine size not published different kind of product: code interpreter machine size per session not published egress price not published	max 200 concurrent egress price not published	None that bite
1,000 short jobs at once	machine size not published different kind of product: code interpreter machine size per session not published egress price not published	max 200 concurrent shared vCPUs on this tier: heavy, sustained CPU use can be slowed by other tenants egress price not published	max 2,400 starts/day (needs ~6,667) max 600 starts/hour (needs ~1,000) bursts above 90 starts/min are queued (a burst of 333 takes ~4 min to start)

Which one should you use?

Use Together Code Sandbox when

you already run models on Together and want light sandboxes, from a tenth of a core, that you save with their memory and branch from, or a simple Python runner billed per session.

Memory snapshots to branch from. For rollouts and test variants: save a sandbox with its memory and start many copies with processes still running.
Sandboxes next to your models. Handy if you already run inference or fine-tuning on Together: sandboxes and models share one account and API key.
Very small machines and a Python runner. For many light jobs: sandboxes start at a tenth of a core, and a Python interpreter API bills per session.

Use Agent 37 when

you host one mostly idle assistant per customer in the US that should sleep with its memory kept and wake on the next message.

Sleeps when idle and keeps its memory. For assistants that sit idle most of the day: an instance sleeps, keeps its memory and wakes on the next message.
Self-serve, with agents ready to run. You can sign up today and start Hermes, OpenClaw, Claude Code or Codex from a template, then chat with it over an API.
Your own image, dedicated cores, US servers. For teams with a Dockerfile and US users: bring your own image, and pick a Performance type on dedicated cores.

Where bochka fits

Look at bochka if your agents do hours of real work, not short runs or idle chat. Together needs you to ask for access and keeps its self-serve product to Python sessions, and Agent 37's default sandbox cannot run Docker. bochka gives every agent a full VM you can start today.

A whole machine on every sandbox

Agents start databases, run Docker and boot emulators. Together documents neither Docker inside nor /dev/kvm, and Agent 37 needs its VM type at 2.5x compute, still with no systemd. bochka has all of it on every sandbox.

Fork a machine while it runs

To try many paths from one half-done task, Together stops the parent and starts copies from its snapshot, and Agent 37 has no fork at all. bochka copies a running sandbox and the original keeps going.

Private hosting and keys kept outside

On Together every port is public and preview URLs are not out yet; on Agent 37 your own API keys sit inside the instance. bochka serves token-protected HTTPS on 50 ports and keeps keys for any HTTPS API in a proxy.

Use bochka when your agents do real work for hours: installs, builds, Docker and emulators on a full VM, with forks of running machines, private hosting and keys kept outside for any API. Compare it one to one: bochka vs Together Code Sandbox and bochka vs Agent 37.

Questions people ask

Is Together Code Sandbox cheaper than Agent 37?

Only for short runs, and Together's priced product, Code Interpreter, does not fit any scenario because sessions end after an hour and keep no disk. For one agent that never stops, Agent 37 is $12 a month against Together's $22. For 50 agents in parallel Together is $264 against $329 on Agent 37's VM type.

Which is faster, Together Code Sandbox or Agent 37?

Neither has a comparable CPU measurement. Agent 37's default instances share host resources, and its Performance type runs on dedicated cores at 4x the compute rate. bochka did 26.69 runs per second on the StarSling Node.js test at 4 vCPU / 8 GB.

Can Together Code Sandbox or Agent 37 run Docker?

Agent 37 runs Docker and Compose on its VM type, but its default gVisor type cannot run a Docker daemon. Together builds sandboxes from Docker images, but does not document Docker inside a sandbox. bochka runs Docker in every sandbox.

Can I sign up for Together Code Sandbox today?

Not on your own: the together-sandbox SDK and CLI must be enabled for your organization by Together. Only Code Interpreter is self-serve. Agent 37 and bochka both let you sign up and start a sandbox right away.

Which one keeps memory when a sandbox stops?

Both can. Together saves disk and memory as a snapshot when a sandbox ends, and Agent 37 checkpoints memory when an instance auto-sleeps. bochka keeps the disk and reboots on resume.

Where do they run?

Agent 37 runs on dedicated servers in the United States. Together does not publish regions for Code Sandbox. bochka runs only in the EU, in Germany, Finland and France, with EU data residency.

Try bochka

Install the CLI, then run bochka new.

Product preview