Workers
Your code answering HTTP: Node, Python or a static Linux binary, or a static site served from the CDN with no code at all. Every deploy is immutable, with instant rollout and rollback.
Built for humans and AI agents alike
Workers, Sandboxes, static sites and data in one platform. One API, one command line tool and one MCP endpoint run all of it, and you pay per org, never per seat.
Free plan, no card. Pro is $20 a month for the whole org.
# Ship a Worker. Every deploy gets its own URL.
$ skopik deploy --runtime python
production -> dep_7f3a… https://api-k2x….skopik.app
deployed in 0.8s
# Give an agent a Linux machine and run code in it.
$ skopik create_sandbox --vcpus 1 --memory-mib 512
$ skopik exec_in_sandbox --sandbox-id $SANDBOX \
--command "python3 -c 'print(6*7)'"
42
# The control API behind everything.
$ curl -X POST $SKOPIK_API/v1/sandboxes \
-H "Authorization: Bearer $SKOPIK_API_KEY" \
-d '{"vcpus": 1, "memory_mib": 512}'
$ curl -X POST $SKOPIK_API/v1/sandboxes/$SANDBOX/exec \
-H "Authorization: Bearer $SKOPIK_API_KEY" \
-d '{"command": "python3 -c \"print(6*7)\""}'
{"exit_code": 0, "stdout": "42\n", …}
// One remote MCP endpoint, one tool per operation.
{ "tool": "create_sandbox",
"arguments": { "vcpus": 1, "memory_mib": 512 } }
{ "tool": "exec_in_sandbox",
"arguments": {
"sandbox_id": "$SANDBOX",
"command": "python3 -c 'print(6*7)'" } }
{"exit_code": 0, "stdout": "42\n", …}
Products
Eight products that know about each other. A Worker declares the data it uses in skopik.json, and the platform hands it access with no keys in your code.
Your code answering HTTP: Node, Python or a static Linux binary, or a static site served from the CDN with no code at all. Every deploy is immutable, with instant rollout and rollback.
Full Linux machines that agents create, run commands in and throw away. Python, Node, git and a C toolchain inside, with a disk that lasts until you destroy it.
A named SQLite database with a single writer. Every write is stored durably before it is acknowledged.
JSON records you can query by any field, and rewind to any moment in the last 30 days. Undo the rewind too.
File storage for objects up to 5 GiB that Workers and Sandboxes call, or mount as a folder.
Job queues that deliver to a Worker and retry failures, with dead letters and cron schedules.
Live publish and subscribe: publish over HTTP, subscribe over WebSocket, with wildcard paths.
Your own domain on any Worker or site, with certificates issued for you.
The console, skopik and the MCP endpoint all call the same API, so a resource made in one is there in the others.
Built for agents
Every operation lives in the control API first, then everywhere else. So anything you can click, an agent can do, with the same words and the same errors.
skopik, curl and MCP equivalent, ready to copy.Shipping
git push to a URL.The developer essentials are there on day one, not on a roadmap.
A GitHub Action deploys the default branch to production and every other branch to its own alias.
Every alias gets its own hostname that follows it, so each pull request has a link to share.
Aliases point at immutable deployments. skopik rollback moves the pointer back.
Tail a Worker's output live with skopik logs, and see requests and status by deployment.
Named values set on a Worker, encrypted with a managed key, never shown again once set.
Put a schedule on a Queue and its Worker gets the job on time, retries included.
Pricing
Bring the whole team and every agent. Both plans include unlimited members.
$0
For trying things and side projects. Hard limits, so it never bills.
$20/ month
Usage included each month, then pay for what you use, up to a spend cap you set.
Compute is GB-hours of running Workers and Sandboxes together. Storage covers Buckets, Shared Objects and Collections. The spend cap starts at $100 of usage past what Pro includes, and you can change it.