Sandbox for background agents.
- Tweets 92
- Following 2
- Followers 10
- Likes 12
Most "long-running" sandboxes live minutes. Some, days.
That is not a background agent. It is a short-lived sandbox with a long life.
The real distinction is not runtime. It is whether the rootfs grows. An agent holding one workspace for months builds a filesystem no fresh sandbox has.
1
7
That 'thinks hard, then waits' shape is why idle has to be cheap. If pausing a waiting agent resets its workspace, it can never accumulate anything across a long task.
We treat the rootfs as durable and compute as replaceable, so waiting costs storage, not a running machine.
5
Selling expensive is a skill. Selling cheap is a harder one.
Same 2 vCPU / 4 GiB sandbox, public list rates, normalized to 1 GiB running for 730 hours:
Sandbox0 $10.95
E2B $30.22
Daytona $30.22
Modal $43.43
Checked 2026-09-23.
2
1
20
Second reason: live migration drains lightly-occupied nodes with process and memory intact, so a handful of background agents don't have to pin a whole machine and inflate the price.
Receipts: sandbox0.ai/pricing
3
Two reasons it holds up.
We bill memory-time only: $0.015/GiB-hour, per second. No CPU line item. Paused sandboxes stop compute charges and keep only storage at $0.02/GiB-month.
No plan fee either.
11
Background agents run for months. Their rootfs only grows.
If pause copies the whole disk, it gets slower every day.
Sandbox0 pauses on a block-COW rootfs: it seals only what changed since the last checkpoint.
Pause scales with change, not size.
1
12
Open source (Apache-2.0). The block-COW rootfs and incremental checkpointing live here:
github.com/sandbox0-ai/sandb…
8
The costly part of autoscaling isn't scaling up.
It's the aftermath. A spike spins up nodes, the spike fades, and you're left with a fleet where each node holds one or two sandboxes — all still billed as running memory.
Sandbox0 now fixes that while workloads keep running.
4
22
The point is consolidation: stop paying for fragmentation.
Compute is $0.015 per GiB-hour of running sandbox memory. Fewer half-empty nodes means less of it.
Normalized to 1 GiB-month, Sandbox0 is $10.95 vs E2B/Daytona $30.22.
Write-up:
sandbox0.ai/blog/2026-09/liv…
11
Measured: a Python workload holding 128 MiB of random memory.
Drain trigger to a usable destination: 3.40-3.64s over six migrations, 3.52s median. None under 3 seconds.
After each move we verified guest PID, memory SHA-256, and an unlinked open file's offset.
7
Files were never the hard part.
A running workspace also holds in-memory state, live subprocesses, and open files with no pathname left. On stock gVisor we checkpoint execution state, transfer it to a compatible node, and resume the same process — same guest PID, same memory.
4
Sandbox0 can live-migrate a running sandbox from one node to another.
It's a platform capability, not something you wire up: no destination to pick, no migration API to call. The scheduler uses it during node evacuation to repack work onto fewer machines.
23
Sandbox0 is offering sandbox credits for this activity. Everyone is welcome to join!
We're teaming up with @Lovable to host the World's FIRST Zero Human Company Hackathon
We're challenging you to build an agent that can run a company completely on its own and earn actual revenue
200 of the most cracked founders, engineers, and students, on August 15th in San Francisco.
Thousands in cash prizes, credits, swag, and good eats.
Big thanks to our partners @Lovable, @render, @thelinqapp, @fastinoAI, @superserve_ai, @replayio, @band_hq, @dodopayments, @svangel, @SignalFire, @solofounders, @Sandbox0AI, and many others to be announced soon
Apps close soon; RSVP in the comments
(interested in sponsoring? DM me!)
2
48
Claude Code SDK uses about 5 times more memory than the Codex SDK, which is worth paying attention to when running large-scale agents.
#codex #claudecode #managedagents
1
60
Every AI agent that calls an external API needs credentials. GitHub tokens, OpenAI keys, database passwords — something has to authenticate the request. The question isn't whether your agent needs credentials. It's whether the agent itself needs to hold them.
sandbox0.ai/blog/2026-03/kee…
28
Some AI agents need to run 24/7 and learn over time:
- Customer service agents learning user patterns
- Monitoring agents accumulating historical data
- Research agents building knowledge bases
For these production workloads, agents need persistent storage across sessions.
Sandbox0 uses JuiceFS to keep state for days, weeks, or months. Your agent remembers what it learned yesterday, last week, last month.
Works alongside fast-start sandboxes for a complete agent infrastructure stack.
1
2
39
Long-running agents need more than fast code execution. They need to remember.
Checkpoint their work, resume from any state, fork to explore branches. Persistent storage transforms stateless sandboxes into workspace with memory.
github.com/sandbox0-ai/sandb…
1
2
42