{ "features_resolved": { "snapshot_any": true, "snapshot_mem": false, "fork": true, "pause_resume": true, "persistent_disk": true, "whole_disk": null, "persist_scope": null, "volumes": true, "auto_stop_idle": false, "long_sessions": true, "custom_image": true, "image_snapshot": true, "image_oci": false, "vm_isolation": true, "docker_inside": true, "nested_virt": false, "root": null, "systemd": null, "browser": true, "desktop": true, "computer_use": false, "code_interpreter": null, "browser_control": null, "desktop_control": null, "browser_and_desktop": false, "anti_bot": false, "captcha_solving": false, "residential_ip": false, "agent_harnesses": true, "ssh": true, "public_ipv4": false, "inbound_https": true, "custom_domain": null, "raw_tcp_inbound": null, "egress_allowlist": false, "open_internet": true, "static_egress_ip": false, "private_network": null, "self_host": false, "byoc": false, "open_source": false, "soc2": null, "hipaa": null, "sso": null, "gpu": false, "arm64": false, "windows": false, "macos": true, "lang_python": true, "lang_node": true, "lang_go": null, "lang_rust": null, "lang_java": null, "gpu_desktop": null, "credential_injection": null, "egress_http_rules": false, "wake_on_request": false, "live_resize": true, "fork_running": false, "memory_snapshot_fork": false, "webhooks": null, "scheduled_wakeups": null, "audit_logs": null, "eu_data_residency": false, "zero_data_retention": null, "scoped_api_keys": null, "spend_limits": true, "mcp_server": false, "usage_api": true, "snapshot_auto": false, "snapshot_on_demand": true, "harness_api": false, "own_agent_api": false, "fast_boot": null, "strong_isolation": true, "ingress_rules": true, "guest_firewall": true, "secret_proxy": false, "secret_proxy_any": false, "volume_attach": false, "volume_shared": true, "sidecar": null, "fuse": true, "kernel_features": true, "burst": null }, "company": { "id": "agentcomputer", "checked": "2026-09-30", "company": "Agent Computer", "ownership": { "value": "independent", "parent": null, "source": "https://github.com/AgentComputerAI", "quote": "Agent Computer" }, "stage": { "value": "unknown", "source": "", "quote": "", "confidence": "low", "note": "No funding announcement, investor list, accelerator batch page, or press coverage could be found. The GitHub organization (AgentComputerAI) was only created 2026-03-28, consistent with a very young, likely pre-seed/bootstrapped startup, but no credible source confirms a stage or any raised round." }, "accelerators": [], "total_funding_usd": { "value": null, "status": "unknown" }, "last_round": { "type": null, "amount_usd": null, "date": null, "leads": [], "status": "unknown", "source": "", "quote": "" }, "investors": [], "portfolio_listings": [], "revenue": { "value": null, "status": "unknown" }, "employees": { "value": null, "source": "" }, "note": "Search engines (DuckDuckGo, Bing, Google, Startpage, Marginalia) all returned bot-challenge/blocked pages in this environment, so third-party press/funding coverage could not be queried; this assessment relies on the company's own site, GitHub org, and direct checks of YC/Crunchbase URLs (no matching public profile found). The GitHub org AgentComputerAI was created 2026-03-28 with only 7 public repos and 5 followers, no public members listed, consistent with a small/very early-stage team. No evidence of acquisition, subsidiary status, or being a product of a larger company was found; site branding (agentcomputer.ai) presents it as an independent company." }, "id": "agentcomputer", "name": "AgentComputer", "url": "https://agentcomputer.ai/pricing", "category": "agent-sandbox", "isolation": "firecracker", "modes": [ { "key": "on-demand", "label": "On-demand", "flags": [], "os": [ "linux" ], "regions": null, "gpu_only": false, "features": { "isolation": "firecracker" }, "pricing": "resource", "vcpu_h": 0.07, "ram_gib_h": 0.04375, "cpu_basis": "active", "ram_basis": null, "active_floor": 0, "ram_per_vcpu": [ null, null ], "ram_fixed_per_vcpu": null, "vcpu_options": null, "min_vcpu": null, "max_vcpu": null, "max_ram_gib": 64, "sizes": null, "pool_tiers": null, "requires_always_on": false, "always_on_month_per_instance": null, "commit_note": null, "min_billed_seconds": null, "granularity_s": null, "start_fee": null, "boot_overhead_s": null, "multiplier": 1, "windows_vcpu_h": null, "gpu": null, "note": "Pricing page: usage-based, CPU Time $0.07/CPU-h, Memory $0.04375/GB-h. Whether memory is billed on allocation or on use is not published (ram_basis null; engine then bills allocation, the conservative case)." }, { "key": "macos-on-demand", "label": "macOS QEMU/KVM on-demand", "flags": [], "os": [ "macos" ], "regions": null, "features": { "macos": true, "isolation": "vm", "desktop": true, "snapshot": "none", "inbound_https": false }, "pricing": "resource", "vcpu_h": 0.07, "ram_gib_h": 0.04375, "cpu_basis": "active", "ram_basis": null, "min_vcpu": null, "max_vcpu": null, "max_ram_gib": null, "min_billed_seconds": null, "granularity_s": null, "start_fee": null, "note": "Pricing page publishes one product-wide PAYG table but no macOS exception or selectable CPU count. Default macOS is 4 GiB RAM and a 64 GiB guest disk. Linux-only exec, file operations, mounts, published ports, resize and snapshots are unavailable on macOS." } ], "storage": { "included_disk_gib": 0, "disk_gib_month": 0.49859, "disk_billed_when_stopped": true, "snapshot_gib_month": null, "paused_ram_billed": null }, "network": { "egress_free_gib": null, "egress_gib": null, "ipv4_month": null, "ipv4_available": false, "internet": { "default": "full" } }, "plans": [ { "name": "Pay as you go", "fee": 0, "flags": [], "fee_is_credit": false, "included_usd": null, "per_seat": null, "concurrency": null, "max_session_h": null, "max_vcpu": null, "max_ram_gib": null, "sizes_allowed": null, "note": "Unknown limits are not promises of unlimited capacity." }, { "name": "Enterprise", "fee": null, "flags": [ "sales" ], "fee_is_credit": null, "included_usd": null, "per_seat": null, "concurrency": null, "max_session_h": null, "max_vcpu": null, "max_ram_gib": null, "sizes_allowed": null, "note": "Unknown limits are not promises of unlimited capacity." } ], "free": { "monthly_credit": null, "one_time_credit": null }, "features": { "linux": true, "windows": false, "macos": true, "arm64": false, "gpu": false, "snapshot": "fs", "pause_resume": true, "fork": true, "persistent_disk": true, "volumes": null, "auto_stop_idle": false, "max_session_h": null, "custom_image": null, "docker_inside": true, "nested_virt": false, "root": null, "systemd": null, "browser": true, "desktop": true, "computer_use": false, "code_interpreter": null, "agent_harnesses": true, "ssh": true, "public_ipv4": false, "inbound_https": true, "custom_domain": null, "raw_tcp_inbound": null, "egress_allowlist": false, "static_egress_ip": false, "private_network": null, "self_host": false, "byoc": false, "open_source": false, "soc2": null, "hipaa": null, "sso": null, "gpu_types": [], "regions": [], "lang_python": true, "lang_node": true, "lang_go": null, "lang_rust": null, "lang_java": null, "gpu_desktop": null, "credential_injection": null, "egress_http_rules": false, "wake_on_request": false, "live_resize": true, "fork_running": false, "memory_snapshot_fork": false, "webhooks": null, "scheduled_wakeups": null, "audit_logs": null, "eu_data_residency": false, "zero_data_retention": null, "scoped_api_keys": null, "spend_limits": true, "mcp_server": false, "browser_control": null, "desktop_control": null, "anti_bot": false, "captcha_solving": false, "egress_ip_type": "datacenter", "ip_note": "Egress/IP and proxy options not documented.", "isolation": "firecracker", "image_oci": false, "snapshot_on_demand": true, "harness_api": false, "ingress_rules": true, "guest_firewall": true, "secret_proxy": false, "volume_attach": false, "volume_shared": true, "snapshot_auto": false, "own_agent_api": false, "max_session_unlimited": true, "open_internet": true, "usage_api": true }, "perf": { "cpu_runs_s": null, "cold_start_ms": null, "burst_tti_ms": null, "dax_s": null }, "knobs": [], "caveats": [ "Public PAYG pricing says actual resource usage and CPU Time: $0.07/CPU-hour, RAM $0.04375/GB-hour; hot running storage $0.000683/GB-hour, cold stopped storage $0.000027/GB-hour. Exact observed-versus-reserved RAM and CPU minimums unverified.", "Features set to null were not verified. No sandbox was purchased or tested; feature/performance statements are documentation claims. Null maximums are unknown unless a cited note explicitly states unlimited.", "A complete monthly bill requires known snapshot/egress/plan limits; do not silently treat missing ancillary charges as free.", "Storage: hot $0.000683/GB-h while running (=$0.4986/GiB-month), cold $0.000027/GB-h while stopped (=$0.0197/GiB-month, modelled as snapshot_gib_month) per agentcomputer.ai/pricing 2026-09-28.", "Memory billing basis (allocated vs used) not published; priced on allocation. Sprites-style active-RAM billing would cut memory cost ~5-8x.", "Docs (agentcomputer.ai/docs, 2026-09-28): default image \"pre-configured with browser access, SSH\"; VNC via computer open --vnc; up to 64 GiB RAM and 250 GiB storage subject to host capacity. Egress price not published." ], "sources": [ "https://agentcomputer.ai", "https://github.com/rivet-dev/sandbox-agent/blob/bbc195cc3fb5a1dd9cb05d8437442768c511e17e/sdks/typescript/src/providers/agentcomputer.ts", "https://agentcomputer.ai/pricing", "https://agentcomputer.ai/" ], "audit_2026_09_29": { "applied": 6, "source": "checked against the provider's live pricing and docs pages", "reviewed": 4 }, "feature_evidence": { "snapshot_on_demand": { "value": true, "source": "https://agentcomputer.ai/docs/api", "quote": "POST\t/v1/computers//snapshots\tCreate a snapshot. ... POST\t/v1/snapshots//restore\tRestore a snapshot into a new computer.", "note": "CLI 'computer snapshot create' / 'snapshot restore restored-computer'; POST /v1/computers can also 'Create a computer from an image or snapshot' (source.kind=snapshot), so new machines boot from it (fork/template). OpenAPI SnapshotResponse has capture_scope enum [filesystem, computer] and runtime_state enum [absent, included], suggesting some snapshots include runtime (memory) state, but docs do not say so explicitly; computer-host README says 'disk snapshots'. Linux only (not on macOS computers).", "checked": "2026-09-29" }, "harness_api": { "value": false, "source": "https://agentcomputer.ai/docs/cli", "quote": "computer claude-login\tInstall Claude Code credentials onto a computer.\ncomputer codex-login\tInstall Codex credentials onto a computer.", "note": "Only credential installation for Claude Code/Codex plus a generic exec API and ConnectRPC process/PTY gateway. The full public OpenAPI spec has no agent/prompt/conversation endpoints. The separate open-source computer-harness repo runs a local browser 'Agent UI' with Claude, Codex and OpenCode, but that is a UI, not an API to drive them.", "checked": "2026-09-29" }, "egress_allowlist": { "value": false, "source": "https://agentcomputer.ai/docs", "quote": "Network speeds up to ~1 Gb/s down and ~1.2 Gb/s up over shared NAT.", "note": "No egress/firewall/network settings anywhere in docs or the complete OpenAPI spec (CreateComputerRequest has additionalProperties:false and no network field). Host uses nftables internally (computer-host README) but no user-facing control.", "checked": "2026-09-29" }, "egress_http_rules": { "value": false, "source": "https://agentcomputer.ai/openapi.json", "quote": null, "note": "No egress controls of any kind in the public API.", "checked": "2026-09-29" }, "ingress_rules": { "value": true, "source": "https://agentcomputer.ai/docs/computers", "quote": "Ports are public by default: the resulting URL is reachable from the open internet without an access session, which is what you want for webhooks and other unauthenticated callers. Pass --private to require an access session instead", "note": "Per-port publish with visibility public|private; only published ports are exposed; private ports need an access session (short-lived token). Share links for browser/VNC with expiry and revocation. SSH via short-lived certificates. No IP allowlists.", "checked": "2026-09-29" }, "guest_firewall": { "value": true, "source": null, "quote": null, "note": "Firecracker microVM with its own kernel (computer-kernels repo); default user is node, exec accepts a user field; root access not explicitly documented.", "checked": "2026-09-29", "derived": true }, "secret_proxy": { "value": false, "source": "https://agentcomputer.ai/docs/cli", "quote": "computer claude-login\tInstall Claude Code credentials onto a computer.", "note": "Credentials are copied into the machine (claude-login/codex-login, exec env vars, S3 keys passed to FUSE mount). No credential-injecting proxy documented.", "checked": "2026-09-29" }, "volume_attach": { "value": false, "source": "https://agentcomputer.ai/docs/computers", "quote": "The current limits are 2-64 GiB RAM in 1 GiB steps and 5-250 GiB storage in 1 GiB steps.", "note": "Storage is a single resizable disk (actions/resize); no block/network volume resource in the complete OpenAPI spec. Only BYO-bucket FUSE mounts (see volume_shared).", "checked": "2026-09-29" }, "volume_shared": { "value": true, "source": "https://agentcomputer.ai/openapi.json", "quote": "\"summary\": \"Queue a FUSE mount create.\" ... \"kind\": {\"enum\": [\"r2\", \"s3\", \"gcs\", \"webdav\"]", "note": "First-party FUSE mount API (POST /v1/computers/{id}/mounts) of your own R2/S3/GCS/WebDAV bucket with read_only flag and vfs_cache_mode; several computers can mount the same bucket (read-write or read-only). Bucket credentials are passed into the mount (not proxied). Only in OpenAPI/SDK (ComputerMountsClient), not in prose docs; Linux only.", "checked": "2026-09-29" }, "docker_inside": { "value": true, "source": "https://raw.githubusercontent.com/AgentComputerAI/computer-kernels/main/configs/x86_64/6.1.158.config", "quote": "CONFIG_CGROUPS=y / CONFIG_NAMESPACES=y / CONFIG_USER_NS=y / CONFIG_NET_NS=y / CONFIG_OVERLAY_FS=y / CONFIG_BRIDGE=y / CONFIG_VETH=y / CONFIG_NF_NAT=y", "note": "derived: the Firecracker guest kernel config ships cgroups, namespaces, overlayfs, bridge/veth networking and NAT — everything Docker needs — and the default image gives the 'node' user passwordless sudo/root. Docker itself is not preinstalled in the computer-guest Dockerfile, but nothing in the kernel or exec API blocks installing and running it.", "checked": "2026-09-30" }, "nested_virt": { "value": false, "source": "https://raw.githubusercontent.com/AgentComputerAI/computer-kernels/main/configs/x86_64/6.1.158.config", "quote": "CONFIG_KVM_GUEST=y (no CONFIG_KVM=y / CONFIG_KVM_INTEL / CONFIG_KVM_AMD entries in the guest kernel config)", "note": "derived: the guest kernel enables paravirtualized KVM-guest support only; it does not build the KVM hypervisor module, so /dev/kvm is not available inside a Linux computer. macOS computers themselves run as a full QEMU/KVM VM on the host, not exposed as nested virt to the guest.", "checked": "2026-09-30" }, "computer_use": { "value": false, "source": "https://agentcomputer.ai/openapi.json", "quote": "Agent Computer runs on computer-host, a Firecracker-based hypervisor... Every computer boots from the computer-guest image ... pre-configured with browser access, SSH", "note": "not offered: the public API/CLI/SDK expose exec, files, ports, VNC/browser connect and snapshots, but no screenshot+click 'computer use' style action API. VNC (computer open --vnc) gives raw remote-desktop pixels/input only, driven by a human or a third-party computer-use agent, not a first-party API.", "checked": "2026-09-30" }, "anti_bot": { "value": false, "source": "https://agentcomputer.ai/pricing", "quote": "Cloud computer infrastructure for ai agents, provisioned on bare metal hosts.", "note": "not offered: outside the product's scope. Agent Computer is generic VM/desktop infrastructure (Firecracker microVMs plus a browser/desktop image); no stealth/fingerprinting-evasion feature is documented anywhere in the docs, API, or repos.", "checked": "2026-09-30" }, "captcha_solving": { "value": false, "source": "https://agentcomputer.ai/pricing", "quote": "Cloud computer infrastructure for ai agents, provisioned on bare metal hosts.", "note": "not offered: outside the product's scope. No CAPTCHA-solving endpoint or mention exists in the OpenAPI spec, CLI, or docs.", "checked": "2026-09-30" }, "egress_ip_type": { "value": "datacenter", "source": "https://agentcomputer.ai/docs", "quote": "Network speeds up to ~1 Gb/s down and ~1.2 Gb/s up over shared NAT.", "note": "Computers share host NAT egress on bare-metal datacenter hosts; no residential/mobile proxy product is offered and no static IP is documented (public_ipv4_available: false).", "checked": "2026-09-30" }, "snapshot": { "value": "fs", "source": "https://agentcomputer.ai/openapi.json", "quote": "SnapshotResponse: capture_scope enum [filesystem, computer], runtime_state enum [absent, included]", "note": "The computer-host README calls these 'disk snapshots' and the schema's default/documented path is filesystem capture; runtime_state has an 'included' option suggesting memory/process state can sometimes be captured, but no docs page confirms restoring a snapshot resumes running processes, so this is modelled conservatively as filesystem-only ('fs'), medium confidence. Linux only, not available on macOS computers.", "checked": "2026-09-30" }, "snapshot_auto": { "value": false, "source": "https://agentcomputer.ai/docs/computers", "quote": "Use snapshots when you want a saved computer state you can return to later. computer snapshot create my-computer", "note": "not offered: snapshots are explicitly user-triggered via CLI/API (computer snapshot create / POST /v1/computers//snapshots); no documented automatic snapshot-on-stop/sleep behavior.", "checked": "2026-09-30" }, "fork_running": { "value": false, "source": "https://agentcomputer.ai/openapi.json", "quote": "createSnapshot: 'Queue snapshot creation.' / createComputer: 'Create a computer from an image or snapshot.'", "note": "Cloning a computer goes through create-snapshot-then-restore-from-snapshot, an asynchronous two-step process (snapshot creation is queued, 202 accepted); there is no single-call clone-while-running endpoint, so a running machine is not cloned instantaneously without any pause in its own lifecycle.", "checked": "2026-09-30" }, "memory_snapshot_fork": { "value": false, "source": "https://agentcomputer.ai/openapi.json", "quote": "SnapshotResponse.runtime_state enum [absent, included]", "note": "No docs text confirms that restoring a snapshot resumes running processes/memory state (the modelled 'snapshot' field above is fs-only); forking is therefore not documented to preserve running memory state.", "checked": "2026-09-30" }, "wake_on_request": { "value": false, "source": "https://agentcomputer.ai/openapi.json", "quote": "POST /v1/computers//actions/start — Start a stopped computer. POST /v1/computers//actions/ensure-ready — Start the computer if it isn't already running or starting.", "note": "not offered: starting a stopped computer is an explicit API/CLI action; there is no documented mechanism where an inbound request to a published port automatically wakes a stopped computer.", "checked": "2026-09-30" }, "auto_stop_idle": { "value": false, "source": "https://agentcomputer.ai/docs", "quote": "Pay only for what you use: runtime CPU/RAM, hot storage, and cold storage are billed separately.", "note": "not offered: computer state transitions (provisioning/starting/running/updating/stopping/stopped) are driven only by explicit start/stop/ensure-ready API calls; no idle-timeout auto-stop is documented.", "checked": "2026-09-30" }, "agent_harnesses": { "value": true, "source": "https://raw.githubusercontent.com/AgentComputerAI/computer-guest/main/Dockerfile", "quote": "RUN npm install -g @anthropic-ai/claude-code @openai/codex && npm cache clean --force", "note": "The default computer-guest image ships Claude Code and Codex CLIs preinstalled, and the CLI has dedicated 'computer claude-login' / 'computer codex-login' commands to install credentials for them.", "checked": "2026-09-30" }, "own_agent_api": { "value": false, "source": "https://agentcomputer.ai/openapi.json", "quote": "paths: /v1/computers, /v1/computers/{id}/exec, /v1/computers/{id}/snapshots, /v1/computers/{id}/mounts, /v1/computers/{id}/ports ... (no agent/task/prompt endpoints)", "note": "not offered: the public API is purely infrastructure (create/exec/files/ports/snapshots/mounts); there is no first-party task/browser/computer-use agent to drive via API.", "checked": "2026-09-30" }, "mcp_server": { "value": false, "source": "https://api.github.com/orgs/AgentComputerAI/repos", "quote": "computer-guest, torch, computer-host, computer-harness, computer-kernels, computer-nix, computer-cli", "note": "not offered: no MCP-server repo/package is published by AgentComputerAI, and no MCP mention appears anywhere in the docs, llms.txt, or OpenAPI spec.", "checked": "2026-09-30" }, "lang_python": { "value": true, "source": "https://raw.githubusercontent.com/AgentComputerAI/computer-guest/main/Dockerfile", "quote": "python3 python3-pip python3-venv pipx", "note": "Preinstalled in the default computer-guest image.", "checked": "2026-09-30" }, "lang_node": { "value": true, "source": "https://raw.githubusercontent.com/AgentComputerAI/computer-guest/main/Dockerfile", "quote": "deb.nodesource.com/node_22.x ... apt-get install -y ... nodejs", "note": "Node.js 22 is preinstalled in the default computer-guest image.", "checked": "2026-09-30" }, "static_egress_ip": { "value": false, "source": "https://agentcomputer.ai/docs", "quote": "Network speeds up to ~1 Gb/s down and ~1.2 Gb/s up over shared NAT.", "note": "not offered: computers egress through shared host NAT; no static/dedicated egress IP option is documented, and public_ipv4 is explicitly unavailable.", "checked": "2026-09-30" }, "gpu": { "value": false, "source": "https://agentcomputer.ai/pricing", "quote": "Pay-as-you-go pricing based on actual resource usage: CPU Time, Memory, Hot Storage, Cold Storage", "note": "not offered: outside the product's scope. Pricing and the computer-images/size-preset API only cover CPU/RAM/storage on Firecracker microVMs (plus a macOS QEMU/KVM VM); no GPU SKU, image, or size preset exists.", "checked": "2026-09-30" }, "windows": { "value": false, "source": "https://agentcomputer.ai/openapi.json", "quote": "CreateComputerRequest.platform enum [linux, macos]; ComputerImageResponse.platform enum [linux, macos]", "note": "The platform enum in the public API only supports linux and macos; there is no Windows option anywhere in the API, CLI, or docs.", "checked": "2026-09-30" }, "arm64": { "value": false, "source": "https://raw.githubusercontent.com/AgentComputerAI/computer-kernels/main/README.md", "quote": "Current bootstrap target: - x86_64 - 6.1.158", "note": "The Firecracker guest kernel build target is x86_64 only; the configs/ directory contains only an x86_64 subfolder.", "checked": "2026-09-30" }, "self_host": { "value": false, "source": "https://agentcomputer.ai/pricing", "quote": "Enterprise ... Private infrastructure options", "note": "not offered as a customer-run product: Agent Computer is sold as a hosted API/CLI service (computer-host/computer-kernels are the provider's own internal infra, not a packaged product customers deploy); Enterprise mentions 'private infrastructure options' via sales, not a documented self-hosting path.", "checked": "2026-09-30" }, "open_source": { "value": false, "source": "https://github.com/AgentComputerAI/computer-cli", "quote": "computer-cli - Public mirror of the Agent Computer CLI", "note": "Only the CLI (public mirror), the guest image, host README, kernel configs, and a Nix harness are open-sourced; the core control-plane/API service itself is closed-source, so the product as a whole is not open source.", "checked": "2026-09-30" }, "eu_data_residency": { "value": false, "source": "https://agentcomputer.ai/openapi.json", "quote": "ComputerCapacityCheckRequest/Response have no region field; no /v1/regions endpoint exists", "note": "not offered: the public API has no region parameter or region-listing endpoint anywhere, and no region/EU mention appears in the docs; the product is described only as bare-metal hosts with no location selection.", "checked": "2026-09-30" }, "max_session_h": { "value": "none", "source": "https://agentcomputer.ai/docs/computers", "quote": "computer power-off my-computer Stop a computer without deleting its stored state. computer power-on my-computer Recreate a computer runtime against the same stored state.", "note": "Computers are billed by actual runtime and only stop when the user calls stop/power-off (or deletes); no maximum session-length limit is documented anywhere in the docs, pricing, or API.", "checked": "2026-09-30" }, "image_oci": { "value": false, "source": "https://agentcomputer.ai/openapi.json", "quote": "CreateComputerRequest.source: {\"kind\": {\"enum\": [\"image\", \"snapshot\"]}, \"image_id\": {\"type\": \"string\"}} ; GET /v1/computer-images -> ListComputerImagesResponse{images: ComputerImageResponse[]}", "note": "Computers can only be created from the provider's own curated computer-images catalog (image_id references a listed ComputerImageResponse, platform enum linux|macos) or from a snapshot of an existing computer. The full OpenAPI spec (additionalProperties:false throughout) has no endpoint to upload/register a custom Docker or OCI image, no registry/tag/digest field, and no 'docker'/'oci' terms anywhere. So bringing your own container/OCI image is not supported; confirms the card's false.", "checked": "2026-09-30" }, "public_ipv4": { "value": false, "source": "https://agentcomputer.ai/docs", "quote": "Network speeds up to ~1 Gb/s down and ~1.2 Gb/s up over shared NAT.", "note": "Docs explicitly describe networking as 'shared NAT', i.e. computers do not get a dedicated public IPv4 address; inbound access is instead via provider-issued hostnames for published ports (docs/computers: 'Ports are public by default: the resulting URL is reachable from the open internet ... derive the final URL from the computer handle and the published port name'), not a raw IP. The complete OpenAPI spec (CreateComputerRequest/ComputerResponse) has no ip/ipv4/address field. Confirms the card's false.", "checked": "2026-09-30" }, "live_resize": { "value": true, "source": "https://agentcomputer.ai/openapi.json", "quote": "POST /v1/computers/{computer_id}/actions/resize — \"Queue a computer resize.\" (202 \"Computer resize accepted.\")", "note": "resize is a computer action (like exec/ports/snapshots) taken against an existing, running computer_id rather than a create-time-only parameter, and docs/computers lists 'resize' among the Linux-only live actions available on a running computer (alongside exec, mounts, published ports). It's an async/queued action (202 accepted) so it may briefly interrupt the guest rather than being instantaneous, but the machine itself is not recreated/deleted.", "checked": "2026-09-30" }, "byoc": { "value": false, "source": "https://agentcomputer.ai/pricing", "quote": "Enterprise ... Custom computer capacity policies, Team billing and managed rollout support, Private infrastructure options, Priority operational support", "note": "not offered: outside the product's scope. Agent Computer describes itself as \"Cloud computer infrastructure for ai agents, provisioned on bare metal hosts\" (its own Firecracker fleet on computer-host); the only enterprise mention is vague sales-gated 'private infrastructure options', with no documentation anywhere (docs, OpenAPI, GitHub org) of deploying the managed control plane into a customer's own AWS/GCP/Azure account.", "checked": "2026-09-30" }, "usage_api": { "value": true, "source": "https://agentcomputer.ai/docs/api", "quote": "GET /v1/computers//usage Read per-computer usage (runtime, CPU-hours, RAM-GiB-hours, hot and cold storage). GET /v1/usage Read account usage.", "note": "granularity: per-sandbox; cost: yes", "confidence": "high", "checked": "2026-10-01" } }, "locations": [], "locations_source": "https://agentcomputer.ai/docs", "locations_quote": "Network speeds up to ~1 Gb/s down and ~1.2 Gb/s up over shared NAT. Computers can be configured up to 64 GiB RAM and 250 GiB storage, subject to currently available host capacity.", "locations_checked": "2026-10-01" }