# OpenComputer

The platform for long-running agents. Define an agent like a serverless function. Choose your model. We handle the runtime.

Source: https://wheredoirunthis.com/products/opencomputer · JSON: https://wheredoirunthis.com/api/products/opencomputer · Read on 2026-10-12

## At a glance

- Type: Sandbox
- Kind: Open core
- Isolation: Virtual machine
- Free tier: Not documented
- Starting price: $20/mo
- GPU: No
- Regions: Not disclosed
- Coverage: 15 of 61 assessed capabilities supported
- Agent readiness: read 11 of 11 signals, drive 9 of 11 signals

## Links

- Homepage: https://opencomputer.dev/
- Docs: https://docs.opencomputer.dev/agents/overview
- Pricing: https://opencomputer.dev/#pricing
- Repository: https://github.com/diggerhq/opencomputer

## Pricing

- Model: Usage-based, one meter set for both layers: (1) model tokens passed through at provider API rates (zero with your own Codex subscription), (2) machine time billed per second while a session/sandbox runs. Plans are PREPAID usage credit with a 10× multiplier; running out drops you to PAYG rates.
- Unit: per-second machine time by memory/vCPU tier (quoted per minute) + per-token model passthrough
- Free tier: $10 free credit to start (PAYG). No perpetual free tier.

| Tier | Price | Includes | Notes |
| --- | --- | --- | --- |
| PAYG | usage-based ($10 free credit) | Model calls at API rates, machine time per second, scale to zero when idle, no monthly commitment. | Burst time billed at peak rate. Pricing table advertises disk maxima by machine baseline; actual hosted lifecycle/capabilities unconfirmed. [source](https://opencomputer.dev/sandboxes) |
| Pro | $20/mo | Prepaid: 'Your $20 becomes 10× the usage credit'; credit covers model calls and machine time; bring your Codex subscription so model calls cost nothing. | Marked 'Popular'. BYOK (Codex account) requires Pro or Max. [source](https://opencomputer.dev/#pricing) |
| Max | $200/mo | 10× prepaid credit at scale — 'Enough to keep 10+ default machines running around the clock'. |  [source](https://opencomputer.dev/#pricing) |
| Enterprise | Custom | Your own cloud or VPC, self-hosted connections, volume pricing and higher limits, SSO, audit log, priority support. | Talk to us. [source](https://opencomputer.dev/#pricing) |

- Overages: When prepaid credit runs out, usage continues at the same PAYG rates. Additional persistent cross-sandbox disks are advertised as forthcoming; no current additional storage/egress charge found.
- Card required: Not documented
- Source: https://opencomputer.dev/#pricing

## Runtime

- **GPU & accelerators**: No GPU offering found on current public pages/docs; not a confirmed runtime prohibition. [source](https://docs.opencomputer.dev/llms-full.txt)
- **BYOC (Bring your own cloud)**: Enterprise tier: 'Your own cloud or VPC, self-hosted connections'. Otherwise self-host the OSS core in your own AWS/Azure/GCP account. — "Enterprise card: 'Your own cloud or VPC, self-hosted connections · Volume pricing and higher limits · SSO, audit log, priority support · Talk to us'." (source: homepage pricing) (confidence: medium)
- **Secrets management**: Managed agent connections use write-only origin/method/path-scoped secret injection, outside prompts/source/runtime; separately configurable runtime variables are encrypted at rest but visible to agent code/tools/processes. Bare-sandbox proxy placement/behaviour is unconfirmed because its docs return 404. [source](https://docs.opencomputer.dev/agents/secrets)
- **Storage options**: Hosted sandbox disk guarantees unknown; advertised max disk 8/16/32 GB by baseline size — "Current pricing table advertises max disk 8 GB for 0.5/1/2 GB baselines, 16 GB for 4 GB baseline, 32 GB for 8 GB baseline. Cross-sandbox disks are "on the way". Managed agents additionally have project SQL and document memory. Actual hosted disk/lifecycle limits not confirmed." [source](https://opencomputer.dev/sandboxes) (confidence: low)
- **Supported workloads**: managed TypeScript serverless AI agents with durable sessions, scheduled/webhook/Slack/GitHub-triggered agent tasks, BYO-harness Linux sandboxes, code execution and coding agents, browser automation on a Linux machine — "Current homepage: "clone a repo, run ffmpeg, drive a browser, install anything"; bare-sandbox page explicitly supports BYO harness." (source: homepage + /sandboxes + docs examples) (confidence: high)
- **Persistence**: unknown — "Hosted sandbox lifetime and disk durability cannot be confirmed because sandbox docs return 404. Landing advertises long-running VMs and says additional cross-sandbox disks are "on the way". Agent sessions, project database and document memory have separately documented durability." [source](https://docs.opencomputer.dev/sandboxes) (confidence: low)
- **Data jurisdiction**: unknown — "No current hosted region inventory or residency statement. Removed sandbox API examples are not current region evidence." (source: current public landing/docs) (confidence: low)
- **Resource limits & quotas**: Hosted implementation limits unknown; advertised baselines/burst sizes and disk maxima recorded — "Sandbox docs 404. Marketing table: 0.5–8 GB baseline, burst 2–32 GB, max disk 8–32 GB. Homepage agent default 2 GB/1 vCPU. New agent projects have one default environment; legacy projects retain development/production. Do not carry forward former 8h, fixed-16GB, no-fork/no-resize claims as current hosting facts." [source](https://opencomputer.dev/sandboxes) (confidence: low)
- **SDK languages**: TypeScript (@opencomputer/sdk 2.3.0; sandbox root and portable /agents client; @opencomputer/agent authoring) and Python (opencomputer-sdk 1.1.0 sandbox SDK). Agent authoring is TypeScript. [source](https://github.com/diggerhq/opencomputer/blob/main/sdks/typescript/README.md)
- **Network & egress controls**: Managed outbound connections constrain credentialed requests; general sandbox firewall unknown — ""OpenComputer checks the declared origin, path prefix, method, agent, and environment before sending the outbound request." Homepage advertises full egress. This is credential-destination isolation, not a general network firewall." [source](https://docs.opencomputer.dev/agents/secrets) (confidence: medium)
- **Snapshots & forks**: Advertised checkpoints/fork; hosted semantics unknown — "Landing: "full Linux microVMs with checkpoint, fork, and live resize". Current docs no longer specify filesystem versus memory capture or hosted fork availability. QEMU self-host core has checkpoint/fork code; do not conflate with hosting." [source](https://opencomputer.dev/sandboxes) (confidence: low)
- **Deployment methods**: cli, sdk, api, template — "Managed agents: opencomputer deploy, management API, linked GitHub starter template. Sandboxes: official SDK and oc CLI. Git starter/import is not Git-push auto-deployment." (source: docs + homepage) (confidence: high)
- **Isolation**: Virtual machine — "Advertised "full Linux microVMs"; documented self-host core uses QEMU/KVM. AWS Lambda MicroVM backend also exists, but hosted VMM/rollout is not established. Do not assert Firecracker." [source](https://github.com/diggerhq/opencomputer/blob/main/SELFHOSTING.md) (confidence: medium)

## Capabilities

### AI & agents

| Capability | Status | Evidence |
| --- | --- | --- |
| Agent skills | Supported | "Package reusable instructions and supporting resources with an agent" [source](https://docs.opencomputer.dev/agents/skills) |
| AI app builder | Not supported | "Define an agent like a serverless function." [source](https://opencomputer.dev/) |
| Autonomous QA agent | Not supported | "Add high-value regression tests" [source](https://docs.opencomputer.dev/agents/examples/test-coverage) |
| Built-in coding agent | Not supported | "OpenComputer owns the agent harness." [source](https://docs.opencomputer.dev/agents/mental-model) |
| BYOK for LLMs | Partially supported | "BYOK lets a Serverless Agent use an organization-connected Codex account" [source](https://docs.opencomputer.dev/agents/byok) |
| Credential-injection proxy | Supported | "The credential is attached only after the destination, method, path," [source](https://docs.opencomputer.dev/agents/secrets) |
| LLM gateway | Supported | "The framework-neutral model gateway for hosted OpenComputer agent Workers." [source](https://github.com/diggerhq/opencomputer/tree/main/cloudflare-workers/oc-gateway) |
| MCP client | Supported | "service. Declare the server once with `defineMcpServer()`, then attach it to an" [source](https://docs.opencomputer.dev/agents/mcp) |
| MCP-server hosting | Partially supported | "Your tools and MCP servers run on that machine - clone a repo, run ffmpeg, drive a browser, install anything." [source](https://opencomputer.dev/) |
| MCP server | Not supported | "MCP (Model Context Protocol) connects an agent to tools hosted by an external" [source](https://docs.opencomputer.dev/agents/mcp) |
| Model inference API (first-party) | Not supported | "Model calls at API rates" [source](https://opencomputer.dev/) |
| Preinstalled coding agent | Not supported | "Define an agent like a serverless function. Choose your model. We handle the runtime." [source](https://opencomputer.dev/) |

### Networking & delivery

| Capability | Status | Evidence |
| --- | --- | --- |
| Automatic HTTPS | Partially supported | "async createPreviewURL(opts: { port: number; domain?: string; authConfig?: Record<string, unknown> }): Promise<PreviewURLResult> {" [source](https://github.com/diggerhq/opencomputer/blob/main/sdks/typescript/src/sandbox.ts) |
| Custom domains | Partially supported | "async createPreviewURL(opts: { port: number; domain?: string; authConfig?: Record<string, unknown> }): Promise<PreviewURLResult> {" [source](https://github.com/diggerhq/opencomputer/blob/main/sdks/typescript/src/sandbox.ts) |
| Direct public IPv6 | Not supported | "- Expose sandbox ports via preview URLs (`oc preview create`)" [source](https://github.com/diggerhq/opencomputer/blob/main/skills/opencomputer/README.md) |
| Edge KV store | Not supported | "OpenComputer is a managed runtime for agents defined as code." [source](https://docs.opencomputer.dev/agents/overview) |
| Egress controls | Partially supported | "Only a relative path is accepted." [source](https://docs.opencomputer.dev/agents/secrets) |
| Email send / receive | Partially supported | "Call Google services, GitHub, and Linear with a credential the platform holds and refreshes" [source](https://docs.opencomputer.dev/agents/services) |
| Public preview URLs | Partially supported | "- Expose sandbox ports via preview URLs (`oc preview create`)" [source](https://github.com/diggerhq/opencomputer/blob/main/skills/opencomputer/README.md) |
| Private networking | Not supported | "full egress" [source](https://opencomputer.dev/) |
| Built-in reverse proxy | Partially supported | "The server owns API," [source](https://github.com/diggerhq/opencomputer/blob/main/SELFHOSTING.md) |

### Organisation & security

| Capability | Status | Evidence |
| --- | --- | --- |
| Built-in auth | Partially supported | "previewAuthToken" [source](https://github.com/diggerhq/opencomputer/blob/main/sdks/typescript/src/sandbox.ts) |
| BYOC | Partially supported | "Your own cloud or VPC, self-hosted connections" [source](https://opencomputer.dev/#pricing) |
| Deploy into customers' VPCs | Not supported | "Define an agent like a serverless function. Choose your model. We handle the runtime." [source](https://opencomputer.dev/) |
| Secrets management | Supported | "OpenComputer secrets are write-only values used by declared outbound" [source](https://docs.opencomputer.dev/agents/secrets) |
| Self-host / OSS core | Supported | "This document describes the supported shape for running OpenComputer in your own cloud account." [source](https://github.com/diggerhq/opencomputer/blob/main/SELFHOSTING.md) |
| Sharing | Partially supported | "- Expose sandbox ports via preview URLs (`oc preview create`)" [source](https://github.com/diggerhq/opencomputer/blob/main/skills/opencomputer/README.md) |
| SSO / SAML | Partially supported | "SSO, audit log, priority support" [source](https://opencomputer.dev/#pricing) |
| Teams | Partially supported | "Only an organization admin can connect or replace the organization account." [source](https://docs.opencomputer.dev/agents/byok) |
| Workload identity | Not supported | "Define an agent like a serverless function. Choose your model. We handle the runtime." [source](https://opencomputer.dev/) |

### Developer access

| Capability | Status | Evidence |
| --- | --- | --- |
| CLI | Supported | "- Create and manage cloud sandboxes (`oc create`, `oc sandbox list`, `oc sandbox kill`)" [source](https://github.com/diggerhq/opencomputer/blob/main/skills/opencomputer/README.md) |
| E2B-compatible API | Not supported | "npm install @opencomputer/sdk" [source](https://github.com/diggerhq/opencomputer/blob/main/sdks/typescript/README.md) |
| Filesystem API | Supported | "await sandbox.files.write("/tmp/test.txt", "Hello, world!");" [source](https://github.com/diggerhq/opencomputer/blob/main/sdks/typescript/README.md) |
| Git deploy | Partially supported | "https://app.opencomputer.dev/new?repository-url=https%3A%2F%2Fgithub.com%2Fdiggerhq%2Fopencomputer-example-hello-world" [source](https://opencomputer.dev/) |
| HTTP data API | Not supported | "Every project session receives a `database` MCP server automatically." [source](https://docs.opencomputer.dev/agents/database) |
| OpenAPI spec | Not supported | "## Errors" [source](https://docs.opencomputer.dev/agents/api) |
| Private container registry | Not supported | "Use the sandbox directly." [source](https://opencomputer.dev/sandboxes) |
| Public API | Supported | "Create sessions, send turns, read the event log and manage projects over HTTP" [source](https://docs.opencomputer.dev/agents/api) |
| SDK | Supported | "The official TypeScript SDK for [OpenComputer](https://github.com/diggerhq/opencomputer): the **cloud sandbox** client at the package root, and the **Serverless Agents** management client on the portable `@opencomputer/sdk/agents` subpath." [source](https://github.com/diggerhq/opencomputer/blob/main/sdks/typescript/README.md) |
| SSH access | Partially supported | "- Open interactive shells (`oc shell <id>`)" [source](https://github.com/diggerhq/opencomputer/blob/main/skills/opencomputer/README.md) |

### Compute & storage

| Capability | Status | Evidence |
| --- | --- | --- |
| Cloud-init / ignition config | Not supported | "const sandbox = await Sandbox.create({ template: "base" });" [source](https://github.com/diggerhq/opencomputer/blob/main/sdks/typescript/README.md) |
| Desktop / computer-use sandbox | Not supported | "drive a browser" [source](https://opencomputer.dev/) |
| GPU compute | Not supported | "Machine time only." [source](https://opencomputer.dev/sandboxes) |
| Live-state (memory) snapshots | Partially supported | "hibernate when idle, resume where they left off." [source](https://opencomputer.dev/) |
| Multi-region | Not supported | "One command. A real computer for every session." [source](https://opencomputer.dev/) |
| Object storage | Not supported | "Additional disks — bottomless storage that persists across sandboxes — are on the way." [source](https://opencomputer.dev/sandboxes) |
| Persistent disk | Partially supported | "Additional disks — bottomless storage that persists across sandboxes — are on the way." [source](https://opencomputer.dev/sandboxes) |
| Sandbox templates / custom images | Partially supported | "const sandbox = await Sandbox.create({ template: "base" });" [source](https://github.com/diggerhq/opencomputer/blob/main/sdks/typescript/README.md) |
| Scale to zero | Supported | "Scale to zero when idle" [source](https://opencomputer.dev/) |
| Selectable isolation runtime | Not supported | "Define an agent like a serverless function. Choose your model. We handle the runtime." [source](https://opencomputer.dev/) |
| Snapshots / forking | Partially supported | "Agents run on OpenComputer sandboxes: full Linux microVMs with checkpoint, fork, and live resize." [source](https://opencomputer.dev/sandboxes) |

### Data

| Capability | Status | Evidence |
| --- | --- | --- |
| Database branching | Not supported | "A durable SQL database shared by every session in a project environment" [source](https://docs.opencomputer.dev/agents/database) |
| Ephemeral instant DB (no-signup) | Not supported | "OpenComputer provisions the database when you deploy." [source](https://docs.opencomputer.dev/agents/database) |
| First-party database | Supported | "Every project environment has one durable SQL database." [source](https://docs.opencomputer.dev/agents/database) |
| Vector search | Not supported | "`documentMemory` stores text under a resource, project and environment." [source](https://docs.opencomputer.dev/agents/document-memory) |

### Platform & operations

| Capability | Status | Evidence |
| --- | --- | --- |
| Managed CI/CD runners | Not supported | "Bringing your own harness or runtime? Use the sandbox directly. Same compute, you own the loop." [source](https://opencomputer.dev/sandboxes) |
| Built-in observability | Supported | "Inspect agent output and outbound-request activity from the OpenComputer CLI" [source](https://docs.opencomputer.dev/agents/logs) |
| Preview environments | Not supported | "New projects have one environment, `default`." [source](https://docs.opencomputer.dev/agents/deployments) |
| Scheduled tasks | Supported | "Schedules start durable agent sessions at recurring times." [source](https://docs.opencomputer.dev/agents/schedules) |
| Project template marketplace | Partially supported | "Pick a template or example app. Make it yours." [source](https://opencomputer.dev/) |
| Durable workflows | Supported | "Sessions are durable: they stream, can be steered mid-run, hibernate when idle, resume where they left off." [source](https://opencomputer.dev/) |

## Agent readiness

### Read: 11 of 11 signals present

11/11 signals (8 full, 3 partial)

- Llms txt: pass — "opencomputer.dev/llms.txt (9.6 KB) links docs, dashboard, GitHub, docs llms.txt, guides + blog as .md — substantive, though its product/pricing copy is stale (meta.scan_issues)."
- Llms full txt: pass — "docs.opencomputer.dev/llms-full.txt: 280895 bytes, 38 current agent/agents-SDK pages, comprehensive for its current index. Previous sandbox docs disappeared; see scan_issues."
- Robots AI policy: partial — "robots.txt: named Allow for Googlebot/Bingbot/Twitterbot/facebookexternalhit + 'User-agent: * Allow: /'; no AI UA named — silent default-allow."
- JSON LD: partial — "Homepage JSON-LD is a single minimal WebSite object."
- Sitemap: pass — "Current /sitemap.xml contains 32 URLs; robots.txt points to it."
- Meta quality: pass — "title + meta description + full OG + Twitter card present and descriptive."
- Markdown negotiation: pass — "Current /agents/overview with Accept: text/markdown returns HTTP 200, content-type text/markdown; request headers and body retained."
- Content efficiency: pass — "Current homepage readable text 6487 bytes / 69027 HTML bytes = 9.40% (scripts/styles excluded), above 5% full-credit threshold."
- Docs discoverable: pass — "'Docs' in header nav, plain <a> links to docs.opencomputer.dev."
- Pricing machine readable: partial — "Current rendered homepage plans and sandbox machine table are parseable; apex llms.txt still publishes incompatible old tiers, so partial credit retained."
- Anti bot: pass — "No challenge on successfully fetched current landing, agent docs, privacy, blog or gh sources. Sandbox docs return true 404, not an anti-bot wall."

### Drive: 9 of 11 signals present

9/11 signals (6 full, 3 partial)

- Public API: pass — "Current /agents/api documented management REST API; official sandbox SDK source exposes REST endpoints. Removed sandbox API pages are not used."
- Openapi spec: fail — "No linked public spec in current /agents/api docs; docs /openapi.json returns HTTP 404 (fetch-manifest)."
- MCP server: fail — "Current agent docs describe consuming HTTPS MCP servers, not a product-management endpoint. Apex .well-known/mcp remains a homepage shell after foxcrawl retry; no documented alternative found."
- SDK packages: pass — "Official TypeScript (@opencomputer/sdk) and Python (opencomputer-sdk) SDKs + @opencomputer/agent."
- Auth docs: pass — "Current /agents/api: x-api-key authentication; organization keys and project-scoped session access documented."
- Agent self serve creds: partial — "`oc login` / `opencomputer login` browser device flow ('give me the link to sign in') or dashboard keys — a human approves; no programmatic key issuance."
- Webhooks: pass — "Current agent webhooks/event subscriptions documentation and SDK README: signed/retried/redeliverable sandbox lifecycle webhooks (Preview)."
- CLI tool: pass — "Current repository README documents TypeScript opencomputer CLI; skills/opencomputer/README.md documents sandbox oc CLI."
- Error model: pass — "JSON error envelope `{ error: { code, message } }` with stable codes + status table."
- Rate limit docs: partial — "429 'Too many requests; retry after Retry-After' documented; no numeric limits or rate-limit headers."
- Agent config: partial — "Current repository skill README: npx skills add diggerhq/opencomputer. One installable skill surface; apex AGENTS.md returns homepage shell."

## Coding harnesses

- Claude Code (Anthropic) — "Works with Claude Code, Codex, Cursor, and 30+ other AI agents that support the [Agent Skills](https://agentskills.io) standard." [source](https://github.com/diggerhq/opencomputer/blob/main/skills/opencomputer/README.md)
- Codex CLI (OpenAI) — "Works with Claude Code, Codex, Cursor, and 30+ other AI agents that support the [Agent Skills](https://agentskills.io) standard." [source](https://github.com/diggerhq/opencomputer/blob/main/skills/opencomputer/README.md)
- Cursor (Anysphere) — "Works with Claude Code, Codex, Cursor, and 30+ other AI agents that support the [Agent Skills](https://agentskills.io) standard." [source](https://github.com/diggerhq/opencomputer/blob/main/skills/opencomputer/README.md)
- OpenCode (opencode) — "Works with Claude Code, Codex, Cursor, and 30+ other AI agents that support the [Agent Skills](https://agentskills.io) standard." [source](https://github.com/diggerhq/opencomputer/blob/main/skills/opencomputer/README.md)
- OpenClaw (Peter Steinberger) — "Works with Claude Code, Codex, Cursor, and 30+ other AI agents that support the [Agent Skills](https://agentskills.io) standard." [source](https://github.com/diggerhq/opencomputer/blob/main/skills/opencomputer/README.md)

## Compliance & regions

- Regions note: No current hosted region inventory or selector found. Prior sandbox docs/examples are now 404 and not reused as residency evidence. Self-host placement is operator-controlled.

## Open source

- Repository: https://github.com/diggerhq/opencomputer
- License: Apache-2.0
- Stars: 563
- Forks: 65
- Self-host methods: source-build, terraform, compose, other
- Self-host dependencies: postgres, redis, object-storage

## Company & presence

- Headquarters: unknown
- Founded: Digger GitHub org created 2020-09-15; opencomputer repo created 2025-12-09.
- Team size: unknown
- Privacy policy: https://opencomputer.dev/privacy/
- GitHub: https://github.com/diggerhq/opencomputer
- Docs: https://docs.opencomputer.dev/
- Blog: https://opencomputer.dev/blog/
- Founder call (cal.com): https://cal.com/team/digger/opencomputer-founder-chat

## 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.