Atlas / 111 products / 22 agent harnesses tracked
Where do I run this?
The independent spec sheet for places to run code, and the code your agents write. Read from public docs. Every claim links to its evidence.
- Products
- 111
- 62 hosted · 20 open core · 29 OSS
- Capabilities
- 61
- in one registry; older scans cover fewer
- Agent harnesses
- 22
- tracked, documented-to-work only
- Sources cited
- 2933
- distinct public pages our facts link to
What do you want to run?
Atlas pick · Persistent environment
exe.dev
Build apps or SSH into a persistent Linux VM.
- Virtual-machine isolation (Source: exe.dev/)
- 100 GiB persistent disk included, shared across VMs (Source: exe.dev/pricing)
- Personal from $15/mo · up to 50 VMs on shared 2 vCPU / 4 GiB (Source: exe.dev/pricing)
- 5 regions · US UK JP DE AU (Source: exe.dev/docs/regions)
01 · The atlas, by isolation boundary
How deep is the wall
around your code?
One tile per product, placed by the isolation primitive its own docs describe. Shallow bands share more of the host. An ordering, not a security rating.
Language sandbox
shares the host process
0/ 5 products
Container
shares the host kernel
0/ 27 products
Kata / gVisor
hardened or user-space kernel
0/ 12 products
Firecracker microVM
shares only the hypervisor
0/ 16 products
Virtual machine
hypervisor, or nothing at all
0/ 32 products
Not disclosed
the public docs do not say
1/ 19 products
Field notes · Scan log
Recently scanned
A small record of what we’ve been looking into lately.
Field notes · Answer log
Recently answered
What people are looking to run. Open a query to see its saved recommendations.
- A persistent VM my coding agent can SSH into, cheap
- List me sandboxes that I can spin under 1s
- For my company, I want to build and offer sandboxes on top of my k8s infrastructure. What can I use & reuse to build the sandbox stack.
- Run untrusted code from my LLM agent, with snapshots to roll back
- A persistent Linux VM for my Node.js coding agent with SSH access and snapshots, under $25 monthly.