Jobs

A job is one command on its own fresh GPU machine. Runcrate starts the machine, runs your command, streams its output, and stops the machine as soon as the command exits. Billing ends with it. Every job has a maximum runtime, so a job never runs unbounded. You do not have to remember to shut anything down.

How a job runs

Submit

You send a command, a GPU type and a maximum runtime. The job is queued while a machine is requested.

Start

A fresh machine with that GPU launches and runs its setup (starting). If you attached a volume, its files are in /workspace. Some machines install the GPU driver after they boot: the job waits for it, so your command starts with the GPU ready.

Run

Your command runs under bash in /workspace (running). Its stdout and stderr are streamed as logs you can follow live.

Stop

When the command exits, hits its maximum runtime or is cancelled, the machine stops and billing ends. Files under /workspace are saved back to the volume.

While the job runs, its machine also shows in Instances. Stopping it there cancels the job.

Run a job

Ask your agent
See the code

Connect the Runcrate MCP server to Claude Code, Codex, Cursor or any MCP client, then ask in plain words:

Run python train.py on an A4000 for at most 2 hours with my volume training-data, then tell me the final loss.

Fine-tune with ./scripts/finetune.sh on 2 H100s, cap it at 6 hours, pass my HF_TOKEN, and tell me if it fails.

What did my last job print before it stopped? Cancel it if it is still running.

What the agent does with the first prompt:

list_volumes to find the id of training-data.
run_job with gpu: "A4000", command: "python train.py", max_runtime_minutes: 120 and the volume’s storage_id. It gets the job id back.
get_job_logs with follow: true, passing the returned cursor back as after each time, until done is true. It reads the loss from the output.
get_job for the final status, exit code and cost, then answers you.

For the second prompt the agent passes the token in run_job’s env object ({ "HF_TOKEN": "..." }). For the third it finds the job with list_jobs, reads it with get_job_logs, and stops it with cancel_job. run_job also takes region, and workspace and environment to choose where the job runs. See MCP tools for every parameter.

The CLI submits the job and follows its logs until it ends:

runcrate run --gpu A4000 --volume <volume-id> --max-runtime 2h -- python train.py

The same job over the API:

curl -X POST https://www.runcrate.ai/api/v1/jobs \
  -H "Authorization: Bearer $RUNCRATE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "command": "python train.py",
    "gpu_type": "A4000",
    "max_runtime_minutes": 120,
    "storage_id": "<volume-id>",
    "env": { "HF_TOKEN": "hf_..." }
  }'

cursor=0
while :; do
  page=$(curl -s "https://www.runcrate.ai/api/v1/jobs/$JOB_ID/logs?after=$cursor&follow=true" \
    -H "Authorization: Bearer $RUNCRATE_API_KEY")
  echo "$page" | jq -j '.data.chunks[].text'
  cursor=$(echo "$page" | jq '.data.cursor')
  [ "$(echo "$page" | jq '.data.done')" = "true" ] && break
done

curl https://www.runcrate.ai/api/v1/jobs/$JOB_ID \
  -H "Authorization: Bearer $RUNCRATE_API_KEY"

curl -X POST https://www.runcrate.ai/api/v1/jobs/$JOB_ID/cancel \
  -H "Authorization: Bearer $RUNCRATE_API_KEY"

In Python, submit and print the output as it arrives:

import os
import requests

API = "https://www.runcrate.ai/api/v1"
headers = {"Authorization": f"Bearer {os.environ['RUNCRATE_API_KEY']}"}

r = requests.post(f"{API}/jobs", headers=headers, json={
    "command": "python train.py",
    "gpu_type": "A4000",
    "max_runtime_minutes": 120,
    "storage_id": "<volume-id>",
})
r.raise_for_status()
job_id = r.json()["data"]["id"]

cursor = 0
while True:
    # follow=true waits up to 25 s for new output, so no sleep is needed
    r = requests.get(f"{API}/jobs/{job_id}/logs", headers=headers,
                     params={"after": cursor, "follow": "true"}, timeout=60)
    r.raise_for_status()
    logs = r.json()["data"]
    for chunk in logs["chunks"]:
        print(chunk["text"], end="")
    cursor = logs["cursor"]
    if logs["done"]:
        break

job = requests.get(f"{API}/jobs/{job_id}", headers=headers).json()["data"]
print(job["status"], job["exit_code"], job["cost_usd"])

The full request and response shapes are in the API reference.

Statuses

Status	Meaning
`queued`	Accepted. A machine is being requested.
`starting`	The machine is launching and running its setup, waiting for the GPU driver, and pulling the `image` if you set one. None of this counts against the maximum runtime.
`running`	Your command has started. `started_at` is set.
`succeeded`	The command exited with code 0.
`failed`	The command exited non-zero (`exit_code` is set), or the machine failed or was stopped (`exit_code` is null and `error` says why, for example the budget was reached or the balance ran out).
`cancelled`	Cancelled through the API, CLI or MCP, or the machine was stopped from Instances.
`timed_out`	The command hit its maximum runtime.

succeeded, failed, cancelled and timed_out are final: they never change, and ended_at is set on all of them.

Logs

Output is stdout and stderr combined, in order, delivered as chunks. A chunk is a raw fragment of output, not a line: print its text as is.

Follow with a cursor. Each response has a cursor. Pass it back as after to get only what is new. With follow=true, a request waits up to 25 seconds for new output or for the job to end.
Stop on done. done: true means the job has ended and you have read every chunk.
Capped at 10 MB per job. The first 5 MB and the newest output are kept. When the middle is dropped, truncated is true and a marker chunk with seq: null sits where the gap is.
Very fast output. When a command writes faster than its output can be sent, the newest output is sent first and the part in between is dropped the same way, so a job’s final lines are never held back.
Kept 30 days after the job ends.

Results and volumes

Attach a volume with storage_id (CLI: --volume) and its files are in /workspace when your command starts. Write checkpoints, metrics and outputs under /workspace: they are saved back to the volume when the machine stops, whatever the final status. Without a volume, files on the machine are gone when it stops, so print what you need to the logs. The volume must be in the same workspace and environment as the job, and not attached to another machine. See Storage.

Environment variables

Pass secrets such as HF_TOKEN in env. Their values reach your command and are never stored or shown again: the job only reports their names in env_keys.

Containers

Set image to run your command inside a container image instead of on the template directly. /workspace is mounted, and the image is pulled while the job is starting, so the pull does not use up the maximum runtime. The image needs bash or sh. The GPUs are passed into the container only on machines that have the NVIDIA container runtime. On other machines the job log says so and the command runs in the container without GPUs. If your command needs the GPU, leave image out: on the machine itself the GPUs are always available.

Workspaces and environments

Jobs belong to a workspace and an environment, like instances. An API key or agent session sees only the jobs in its own environment. Otherwise, pick one with the X-Environment header or the environment query parameter (MCP: the environment parameter).

Billing

A job is billed at its machine’s hourly rate (cost_per_hour, in dollars) from when the machine is running until it stops. Setup time is included. cost_usd on the job is the cost so far and is final once the job ends. See pricing for current rates per GPU-hour. The maximum runtime is counted from when your command starts, so setup does not use it up.

GPU readiness

Your command starts once the GPU driver answers. On machines that install the driver after they boot, the job stays starting while it waits, for up to 20 minutes. If the driver is still not ready after that, the command runs anyway, the job log says so, and GPU calls in it will usually fail, so the job usually ends failed. Cancel and resubmit if that happens; a new job gets a fresh machine.

Limits

Limit	Value
Maximum runtime	1 minute to 24 hours, required on every job
GPUs per job	1 to 8
Command	64 KB
Environment variables	100 keys, 64 KB in total
Log size	10 MB per job (first 5 MB and newest output kept)
Log retention	30 days after the job ends
Log long-poll	Up to 25 seconds per request

Errors

Code	Status	When
`validation_error`	422	A field is missing or out of range.
`BUDGET_REACHED`	402	The environment’s monthly budget is spent. Checked before any machine is requested.
`insufficient_credits`	402	Your balance cannot cover the job.
`bad_request`	400	The volume cannot be used, for example it is attached to another machine.
`not_found`	404	The job or volume does not exist in this workspace and environment.
`conflict`	409	You cancelled a job that already ended, or another launch claimed the volume first.
`service_unavailable`	503	Jobs are briefly unavailable. Retry shortly.

If the machine cannot be launched after the job was created, the error carries details.job_id and that job is failed.

When to use an instance instead

A job fits work that starts, runs and ends on its own: training runs, evaluations, batch inference, data processing. Use an instance when you need to work on the machine yourself (SSH, Jupyter, VS Code), keep a server up for others to call, or run longer than 24 hours.

Welcome

SDKs

CLI

Inference

Compute

Storage

Dedicated Clusters

Billing

Account

FAQ

Jobs

How a job runs

Run a job

Statuses

Logs

Results and volumes

Environment variables

Containers

Workspaces and environments

Billing

GPU readiness

Limits

Errors

When to use an instance instead

​How a job runs

​Run a job

​Statuses

​Logs

​Results and volumes

​Environment variables

​Containers

​Workspaces and environments

​Billing

​GPU readiness

​Limits

​Errors

​When to use an instance instead

How a job runs

Run a job

Statuses

Logs

Results and volumes

Environment variables

Containers

Workspaces and environments

Billing

GPU readiness

Limits

Errors

When to use an instance instead