# Fast VM

Sandboxing infrastructure for AI — virtual machines that boot in 70ms, zero-downtime snapshots, native branching, 50-90% compute savings

Source: https://wheredoirunthis.com/products/fastvm · JSON: https://wheredoirunthis.com/api/products/fastvm · Read on 2026-06-05

## At a glance

- Type: Sandbox
- Kind: Hosted
- Isolation: Virtual machine
- Free tier: No
- Starting price: Not disclosed
- GPU: No
- Regions: Not disclosed
- Coverage: 10 of 35 assessed capabilities supported · 26 not assessed
- Agent readiness: read 10 of 11 signals, drive 7 of 11 signals

## Links

- Homepage: https://fastvm.org/
- Docs: https://fastvm.org/docs
- Pricing: https://fastvm.org/pricing

## Pricing

- Model: Graduated usage — vCPU-hour billing (rate drops with monthly volume) + volume storage; pausing a VM stops the meter; no subscription
- Unit: vCPU-hour (graduated by monthly volume) + volume GiB-month
- Free tier: No

| Tier | Price | Includes | Notes |
| --- | --- | --- | --- |
| Usage (graduated vCPU-hour) | $0.09 → $0.06 per vCPU-hour (by monthly volume) + $0.09/GiB-mo volume | pure usage, no subscription; pausing a VM stops the meter; machine-readable rate ladder at GET /v1/pricing; No included bundle — metered from the first vCPU-hour. |  [source](https://fastvm.org/pricing) |

- Overages: vCPU-hour $0.09 → $0.06 graduated by monthly volume; volume storage $0.09/GiB-month; pausing stops the meter
- Card required: Not documented
- Source: https://fastvm.org/pricing

## Runtime

- **GPU & accelerators**: No — "No GPU mention anywhere (homepage, docs, OpenAPI, llms-full)." [source](https://fastvm.org/) (confidence: medium)
- **BYOC (Bring your own cloud)**: No — "Hosted. BYO GCS/S3 bucket MOUNTS exist (storage), but not bring-your-own-cloud compute." [source](https://fastvm.org/docs) (confidence: medium)
- **Secrets management**: envVars injected at boot (string→string map; ≤256-byte keys, ≤4096-byte values) — "CreateVMRequest.envVars — environment variables injected into the VM at boot. No dedicated secrets vault." [source](https://fastvm.org/openapi.decorated.yaml) (confidence: medium)
- **Storage options**: diskGiB per VM (overridable at launch); managed Volumes (POSIX-coherent multi-attach via virtio-fs, $0.09/GiB-mo); BYO GCS/S3 bucket mounts; snapshot import from OCI image / Dockerfile — "/v1/volumes (multi-attach), /v1/vms/{id}/bucket-mounts (BYO GCS/S3), /v1/snapshot_imports (Docker/OCI/Dockerfile)." [source](https://fastvm.org/docs) (confidence: high)
- **Supported workloads**: reinforcement-learning, ai-coding-agents, browser-automation, non-deterministic-computer-use — "Homepage use cases: RL post-training (fork into branches), coding agents (parallel attempts), browser automation, non-deterministic computer use (Blender/CAD/Photoshop)." [source](https://fastvm.org/) (confidence: high)
- **Persistence**: VM lifecycle provisioning/running/stopped; pausing a VM stops the meter; state via full-VM snapshots (incl. TCP connections) + managed Volumes (POSIX multi-attach virtio-fs) — "Snapshots capture full VM state; managed Volumes ($0.09/GiB-mo) persist independently; pausing stops billing." [source](https://fastvm.org/docs) (confidence: high)
- **Data jurisdiction**: region 'sfo' referenced in benchmarks; no region selection or data-residency disclosed — "Only 'sfo' (San Francisco) appears, in the boot-time benchmark; no selectable region matrix." [source](https://fastvm.org/) (confidence: low)
- **Resource limits & quotas**: org-level quotas: vcpu, memoryMiB, diskGiB, snapshotCount, volumeCount, volumeGiB (limits + usage); 16 services/VM; exec stdout/stderr capped (~4 MiB, truncation flags); default values require auth to query — "Quota object + GET /v1/org/quotas (auth-gated); per-VM service cap 16." [source](https://fastvm.org/openapi.decorated.yaml) (confidence: medium)
- **SDK languages**: python, typescript — "Python SDK (`fastvm` v0.7.0, PyPI, ≥3.9) + TypeScript SDK (`fastvm`, npm) — both 52 methods (llms.txt)." [source](https://fastvm.org/docs) (confidence: high)
- **Network & egress controls**: Yes — "Per-VM firewall with ingress (default deny) + egress (default allow) + DNS axes; IPv4 & IPv6 source CIDRs accepted; port ranges. Configurable outbound policy." [source](https://fastvm.org/openapi.decorated.yaml) (confidence: high)
- **Snapshots & forks**: Yes — "0ms snapshot pause ('VM keeps running' during capture), native branching with branch restore <100ms, 'fork environments into a hundred branches' (RL). Snapshot import from Docker/OCI image or Dockerfile. (NOTE: the intel's older '1000 VMs in 170ms' claim is no longer on the site.)" [source](https://fastvm.org/) (confidence: high)
- **Deployment methods**: sdk, api, cli, ssh — "REST API (api.fastvm.org), Python+TS SDKs, `python -m fastvm` module CLI, console-token WebSocket TTY, and SSH via a CA-signed gateway (ssh <vmId>@<org-slug>.ssh.<stack-domain>)." [source](https://fastvm.org/docs) (confidence: high)
- **Isolation**: Virtual machine — "Homepage isolation column = 'Dedicated kernel' (per VM); hypervisor UNDISCLOSED (Firecracker/KVM/libkrun never named in homepage, docs, OpenAPI, or llms-full). Clear own-kernel VM with an unnamed hypervisor → `vm`. (A 3rd-party reimplementation aitorpazos/fastestvm uses crun+libkrun, but that is not Fast VM's code.)" [source](https://fastvm.org/) (confidence: medium)

## Capabilities

### AI & agents

| Capability | Status | Evidence |
| --- | --- | --- |
| Agent skills | Not supported | "No installable agent-skills system." [source](https://fastvm.org/) |
| AI app builder | Not supported | "No prompt-to-app builder." [source](https://fastvm.org/) |
| Autonomous QA agent | Not assessed |  |
| Built-in coding agent | Not supported | "No first-party coding agent — Fast VM is where agents run." [source](https://fastvm.org/) |
| BYOK for LLMs | Not applicable | "Not an LLM product." [source](https://fastvm.org/docs) |
| Credential-injection proxy | Not assessed |  |
| LLM gateway | Not supported | "No LLM gateway/router." [source](https://fastvm.org/docs) |
| MCP client | Not supported | "Not an MCP client." [source](https://fastvm.org/docs) |
| MCP-server hosting | Not assessed |  |
| MCP server | Not supported | "No MCP server; /.well-known/mcp 404." [source](https://fastvm.org/.well-known/mcp) |
| Model inference API (first-party) | Not assessed |  |
| Preinstalled coding agent | Not assessed |  |

### Networking & delivery

| Capability | Status | Evidence |
| --- | --- | --- |
| Automatic HTTPS | Supported | "Per-VM HTTP services exposed at https://<name>--<vmid>.proxy.<stack-domain> (auto-TLS via the proxy)." [source](https://fastvm.org/docs) |
| Custom domains | Not supported | "Services live on .proxy.<stack-domain> subdomains; no bring-your-own custom-domain feature documented." [source](https://fastvm.org/docs) |
| Direct public IPv6 | Not assessed |  |
| Edge KV store | Not assessed |  |
| Egress controls | Supported | "Per-VM firewall: ingress (default deny) + egress (default allow) + DNS axes; IPv4/IPv6 CIDRs + port ranges — configurable outbound policy." [source](https://fastvm.org/openapi.decorated.yaml) |
| Email send / receive | Not supported | "No email primitive." [source](https://fastvm.org/docs) |
| Public preview URLs | Supported | "Per-VM service URLs (https://<name>--<vmid>.proxy.<domain>) are shareable public endpoints." [source](https://fastvm.org/docs) |
| Private networking | Not assessed |  |
| Built-in reverse proxy | Supported | "Built-in service proxy (incl. a public 4-to-6 HTTP proxy) routes external traffic to VM ports." [source](https://fastvm.org/docs) |

### Organisation & security

| Capability | Status | Evidence |
| --- | --- | --- |
| Built-in auth | Not supported | "API-key auth for the platform; no auth-as-a-service for workloads." [source](https://fastvm.org/docs) |
| BYOC | Not supported | "Hosted; BYO bucket mounts are storage, not BYO-cloud compute." [source](https://fastvm.org/docs) |
| Deploy into customers' VPCs | Not assessed |  |
| Secrets management | Partially supported | "envVars injected at VM boot (string→string map); no dedicated secrets vault." [source](https://fastvm.org/openapi.decorated.yaml) |
| Self-host / OSS core | Not supported | "Hosted; only SDKs (and a forked eval repo) are public — no self-hostable core. The OpenAPI spec is Apache-2.0 but that is spec-licensing, not a runnable core." [source](https://github.com/fastvm-org) |
| Sharing | Partially supported | "Per-VM service URLs are shareable; no team/invite share-access primitive documented." [source](https://fastvm.org/docs) |
| SSO / SAML | Not supported | "No SSO/SAML found." [source](https://fastvm.org/) |
| Teams | Partially supported | "Org-scoped resources (orgId, GET /v1/org/quotas, org-slug in SSH/proxy hostnames) imply an org/team model; explicit member/role management not documented on the public surface." [source](https://fastvm.org/openapi.decorated.yaml) |
| Workload identity | Not assessed |  |

### Developer access

| Capability | Status | Evidence |
| --- | --- | --- |
| CLI | Partially supported | "`python -m fastvm launch` is a module CLI entry point — not a standalone official CLI binary." [source](https://fastvm.org/docs) |
| E2B-compatible API | Not assessed |  |
| Filesystem API | Not assessed |  |
| Git deploy | Not supported | "No git push-to-deploy (snapshot import is from Docker/OCI/Dockerfile, not a git flow)." [source](https://fastvm.org/docs) |
| HTTP data API | Not assessed |  |
| OpenAPI spec | Supported | "Fetchable OpenAPI 3.1.0 (29 paths / 48 operations), no auth required to read." [source](https://fastvm.org/openapi.decorated.yaml) |
| Private container registry | Not supported | "Imports from Docker/OCI images + BYO GCS/S3 bucket mounts; does not host a private registry." [source](https://fastvm.org/docs) |
| Public API | Supported | "Documented REST API (api.fastvm.org), X-API-Key auth; 48 operations across 14 tags." [source](https://api.fastvm.org/) |
| SDK | Supported | "Official Python (`fastvm`) + TypeScript (`fastvm`) SDKs, 52 methods each." [source](https://fastvm.org/docs) |
| SSH access | Supported | "SSH into a VM via a CA-signed gateway: ssh <vmId>@<org-slug>.ssh.<stack-domain>." [source](https://fastvm.org/docs) |

### Compute & storage

| Capability | Status | Evidence |
| --- | --- | --- |
| Cloud-init / ignition config | Not assessed |  |
| Desktop / computer-use sandbox | Not assessed |  |
| GPU compute | Not supported | "No GPU offering mentioned." [source](https://fastvm.org/) |
| Live-state (memory) snapshots | Not assessed |  |
| Multi-region | Not supported | "Only region 'sfo' referenced (in a benchmark); no multi-region selection disclosed." [source](https://fastvm.org/) |
| Object storage | Not assessed |  |
| Persistent disk | Supported | "Per-VM diskGiB + managed Volumes (POSIX multi-attach, $0.09/GiB-mo) persist independently of VM lifecycle." [source](https://fastvm.org/docs) |
| Sandbox templates / custom images | Not assessed |  |
| Scale to zero | Partially supported | "Pausing a VM stops the meter (no compute billed while paused); no documented auto-idle-suspend / wake-on-request primitive." [source](https://fastvm.org/pricing) |
| Selectable isolation runtime | Not assessed |  |
| Snapshots / forking | Supported | "0ms snapshot pause (VM keeps running during capture), native branching, branch restore <100ms, 'fork environments into a hundred branches'." [source](https://fastvm.org/) |

### Data

| Capability | Status | Evidence |
| --- | --- | --- |
| Database branching | Not assessed |  |
| Ephemeral instant DB (no-signup) | Not assessed |  |
| First-party database | Not supported | "No managed database." [source](https://fastvm.org/docs) |
| Vector search | Not assessed |  |

### Platform & operations

| Capability | Status | Evidence |
| --- | --- | --- |
| Managed CI/CD runners | Not assessed |  |
| Built-in observability | Not assessed |  |
| Preview environments | Not assessed |  |
| Scheduled tasks | Not supported | "No cron/scheduled-task primitive documented." [source](https://fastvm.org/docs) |
| Project template marketplace | Not assessed |  |
| Durable workflows | Not supported | "No durable-execution/workflow engine." [source](https://fastvm.org/docs) |

## Agent readiness

### Read: 10 of 11 signals present

10/11 signals (9 full, 1 partial)

- Llms txt: pass — "GET /llms.txt 200 — structured index linking Python/TS SDK refs, REST API, OpenAPI YAML, llms-full"
- Llms full txt: pass — "GET /llms-full.txt 200 (~130KB) — full SDK + API docs for LLM ingestion"
- Robots AI policy: partial — "robots.txt allows all, disallows console/auth paths; no AI-UA directives → silent default-allow"
- JSON LD: pass — "SoftwareApplication JSON-LD block with name + description + offers"
- Sitemap: pass — "GET /sitemap.xml 200 — 6 URLs (/, /pricing, /blog, /docs/python, /docs/typescript, /docs/api)"
- Meta quality: pass — "Descriptive <title> + complete meta description + og:site_name"
- Markdown negotiation: fail — "Accept: text/markdown returns HTML (Next.js app); no markdown variant"
- Content efficiency: pass — "Next.js SSR — content present in source HTML (not a pure JS shell)"
- Docs discoverable: pass — "Docs first-class in nav (/docs, /docs/python, /docs/typescript, /docs/api)"
- Pricing machine readable: pass — "Structured graduated rate ladder on /pricing + an unauthenticated GET /v1/pricing API"
- Anti bot: pass — "All pages GET 200, no challenge"

### Drive: 7 of 11 signals present

7/11 signals (5 full, 2 partial)

- Public API: pass — "Documented REST API at api.fastvm.org (48 operations, 14 tags)"
- Openapi spec: pass — "Fetchable OpenAPI 3.1.0 at fastvm.org/openapi.decorated.yaml (no auth)"
- MCP server: fail — "No MCP; /.well-known/mcp 404"
- SDK packages: pass — "Official Python + TypeScript SDKs (52 methods each)"
- Auth docs: pass — "X-API-Key documented in OpenAPI securitySchemes (FASTVM_API_KEY)"
- Agent self serve creds: partial — "Signup/console exists; key issuance via console (human step); no documented programmatic key issuance"
- Webhooks: fail — "No webhook feature in spec or docs"
- CLI tool: partial — "`python -m fastvm` module CLI; no standalone official CLI binary"
- Error model: pass — "OpenAPI 3.1 defines response/error schemas; ExecVMResponse {exit_code, stdout, stderr, timed_out, *_truncated, duration_ms} + NDJSON streaming — a complete contract"
- Rate limit docs: fail — "No API rate-limit docs (org resource quotas are capacity, not request limits)"
- Agent config: fail — "No agents.md / .well-known/agent / skills listing (llms.txt counts under parse, not agent_config)"

## Coding harnesses

Plausible compatibility (derived, not stored):

- Aider
- Amp
- Antigravity
- Claude Code
- Codex CLI
- Devin
- Droid
- Gemini CLI
- Goose
- Hermes Agent
- Kilo Code
- Kiro CLI
- OpenCode
- Pi
- Qwen Code

## Compliance & regions

- Regions note: Only region 'sfo' (San Francisco) appears — in the boot-time benchmark — not as a selectable placement. No region matrix or data-residency disclosed → disclosed:false.

## Company & presence

- Headquarters: Unknown
- Founded: Unknown (live API + SDKs + benchmark labeled May 2026)
- Team size: Unknown — opaque; no team page
- GitHub: https://github.com/fastvm-org
- PyPI: https://pypi.org/project/fastvm/
- npm: https://www.npmjs.com/package/fastvm
- Blog: https://fastvm.org/blog

## Nearby in the atlas

- [Agent Safehouse](/products/agent-safehouse.md)
- [Agent Sandbox](/products/agent-sandbox.md)
- [Agent Substrate](/products/agent-substrate.md)
- [Agent37](/products/agent37.md)
- [AgentNest](/products/agentnest.md)
- [Anthropic Sandbox Runtime (srt)](/products/sandbox-runtime.md)

## Method

Facts come from the product's own public pages. Each claim carries its evidence: a quote and the page it came from. We count capabilities rather than grade products, and "Not assessed" means we have not checked, never that the product lacks it.