A computer for your coding agent. Its files are still there tomorrow.
A machine of its own with a home disk at /data that outlives the machine, a desktop you can watch or take over, and Codex, OpenCode, git and Node.js ready. After ten idle minutes a small computer sleeps with its memory and a base or large one shuts down; the home stays.
ry computer create dev-1
Computers, in detail.
Files that stay
The home disk at /data is a volume, kept while the computer is shut down and when its machine is replaced. Keep your work there.
A desktop you can watch
A view link for any number of viewers. A control link for keyboard and mouse: one holder per machine. An agent inside can read that a person holds control. An audit shows who got which link.
Ready to work
Node.js with npm, Codex, OpenCode, git, the GitHub CLI, Python 3 with pip, a browser and a desktop. Inside, rytiv-setup claude runs Anthropic's own installer, with your own Claude account.
Two ways to connect Claude Code
On your laptop with Rytiv as its MCP tool: it creates computers, runs commands and snapshots them. Or inside the computer, installed with rytiv-setup claude.
A shell without SSH keys
ry exec runs one command and exits with its code. ry ssh opens a shell over the API. ry proxy --loopback brings a port to your machine without a public port. Each wakes a shut-down computer.
Snapshots for agent computers
A computer made with the MCP tool computer_create keeps checkpoints outside the host, can be snapshotted before a risky step, and comes back from its last checkpoint after a host failure.
Create it, work in it, watch it.
From your terminal, from the API, or let your agent create its own computer over MCP.
ry computer create dev-1 --size base ry ssh dev-1 ry exec dev-1 --dir /data -- npm test ry proxy dev-1 3000:3000 --loopback ry computer view dev-1 # watch, any number of viewers ry computer control dev-1 --ttl 1800 # keyboard and mouse, one holder ry computer audit dev-1
curl -X POST https://api.rytiv.com/v1/apps \
-H "Authorization: Bearer $RYTIV_TOKEN" \
-H "Content-Type: application/json" \
-d '{"name": "dev-1", "computer": true}'
# then POST /v1/volumes (dev-1-home), POST /v1/apps/dev-1/attach,
# PUT /v1/apps/dev-1/secrets (COMPUTER_VIEW_KEY),
# POST /v1/apps/dev-1/deploy with the computer image and size
curl -X POST https://api.rytiv.com/v1/machines/<id>/viewer \
-H "Authorization: Bearer $RYTIV_TOKEN" \
-H "Content-Type: application/json" \
-d '{"mode": "view", "ttlS": 900}' # once: add Rytiv to Claude Code
claude mcp add --transport http rytiv https://api.rytiv.com/v1/mcp --header "Authorization: Bearer <your key>"
> Create a computer, clone my repository into it, run the tests.
computer_create {"name": "agent-1", "image": "<your image>",
"memoryMb": 4096, "volumeGb": 100}
computer_exec {"app": "agent-1", "command": ["git", "status"]}
computer_snapshot {"app": "agent-1", "label": "before-migration"}
→ machineId, label The bounds, as the platform enforces them.
Limits
- small
- 1 vCPU, 2048 MB, 50 GB home
- base (default)
- 2 vCPU, 4096 MB, 100 GB home
- large
- 4 vCPU, 8192 MB, 128 GB home
- Computers per team
- 2; contact us for more
- Idle shutdown
- after 10 minutes without activity
- View and control links
- 60 to 3600 s, default 900
- Shell sessions
- 16 per machine, 64 per organization
- Name
- 2 to 31 characters
Every bound is on Limits and quotas. A request outside them answers 400 with bad-request and names the field.
Errors you will meet
| 403 | org-limit-computers2 computers in a team; delete one, or contact us |
| 409 | control-heldSomebody else holds control; force takes it |
| 409 | machine-stoppedThe machine is stopped; start it first |
| 409 | several-machinesControl needs exactly one machine |
| 429 | too-many-sessionsThe machine has as many shell sessions open as it may |
Every error is JSON with a sentence and a stable code. Branch on code. See the error catalog.
A base computer: €0.0304 per hour of memory while awake, CPU only while it computes, and €10 per month for its 100 GB home disk. Shut down, it costs only the disk.
For businesses and freelancers. All prices excluding VAT. Prepaid credit in euros, no plans, no seats. Introductory pricing until 31 January 2027. Pricing
What developers ask about Computers.
Answers from the documentation. The Computers guide has the rest.
What survives a shutdown?
Everything under /data. A small computer sleeps with its memory, so its programs come back. A base or large computer boots again with /data in place.
What counts as activity?
CPU load, a held control lease and an open viewer with frames flowing. The next ry exec, ry ssh, request or the Wake button in the console starts it again.
Can I use Claude Code?
Yes. On your laptop with Rytiv as its MCP tool, or inside the computer: rytiv-setup claude runs Anthropic's own installer, and you sign in with your own Claude account.
Can I watch while the agent works?
Yes. ry computer view gives a link any number of people can watch. ry computer control takes keyboard and mouse until you release. An agent inside can read that a person holds control.
Console computer or agent computer?
A computer from the console or ry computer create shuts down when idle and keeps its home. One made with the MCP tool computer_create stays up and can be snapshotted.
What does deleting a computer remove?
ry apps destroy removes the machine. The home volume stays until you delete it too with ry volume destroy. Neither can be undone.
On the same API, the same bill.
Give your agent a computer.
Up to 2 computers per team; contact us for more.
claude mcp add --transport http rytiv https://api.rytiv.com/v1/mcp --header "Authorization: Bearer <your key>"
For businesses and freelancers, from a solo developer to a team. Pricing