list-collectionsLists all collections.
LABSCO SUMMARY
Run and check Jetty work without leaving the chat.
The same tools work in Cursor, VS Code, Codex and inside Jetty's own plugin, so your assistant can go from listing your tasks to starting a run, reading the result and putting the task on a schedule, all from the conversation.
What it is
Jetty runs AI tasks in a managed sandbox and scores each result against checks you set; a task is written as a runbook, one markdown file with the job, what counts as done, and the checks. This is Jetty's own MCP server. Your editor starts it on your computer with npx, and it calls Jetty with your token. Jetty has no hosted address for it.
What you get
- Collections and tasks listed and read, and tasks created or updated from a workflow definition
- Runs started in the background, or run to the end in one call when you want the result in the chat (30 to 60 seconds is typical)
- The full record and statistics of past runs, and labels added to a run
- The library of step templates, with each one's inputs
- A check of the keys a runbook needs against what the collection has, and environment variables set without opening the dashboard
- Schedules created, changed, paused, resumed, fired at once or deleted, and their recent runs listed
- Jetty's free-trial keys checked and switched on for a collection
Requirements
- A Jetty account and an API token, created under Settings, API keys; it starts with mlc_.
- The token goes in JETTY_API_TOKEN. Without it, the server reads ~/.config/jetty/token, which the jetty login command or the Jetty plugin's setup wizard writes.
- Each token works for one collection; a request against another collection is refused.
- Node.js 18 or later.
- A Jetty plan: the free plan includes 12 runs a month, with no card needed; paid plans start at $599 a month.
Setup effort
One command plus a key — claude mcp add jetty -- npx -y jetty-mcp-server, then supply credentials
USE CASES
01
Running a task from the chat
List the collections, open the task, and start a run: in the background for long jobs, or waiting for the result when it takes a minute or less.
02
Finding out why a run went wrong
Pull up the task's recent runs, read the full record of the one that failed, and label it.
03
Getting keys in place before the first run
Compare the keys a runbook needs with what the collection has, then set the missing ones without opening the dashboard.
04
Putting a task on a schedule
Create a schedule for a task, fire it once straight away to check it, then pause or resume it as needed.
05
Building a new task from existing steps
Browse the step templates and their inputs, then create the task from a workflow definition.
06
Trying Jetty before adding your own AI keys
Check whether a collection can use Jetty's free-trial keys, then switch them on.
Explore all 27 tools
Named in the project's own documentation.
get-collectionGets one collection and the names of its environment variables.
list-tasksLists the tasks in a collection.
get-taskGets one task and its workflow definition.
create-taskCreates a task from a workflow definition.
update-taskChanges the workflow or description of a task.
get-trial-statusGets the trial-key status of a collection.
activate-trialActivates trial keys for a collection.
run-workflowStarts a run and returns its workflow_id immediately.
run-workflow-syncStarts a run and waits for the result (30 to 60 seconds is typical).
list-trajectoriesLists the recent runs of a task.
get-trajectoryGets the full record of one run.
get-statsGets the run statistics of a task.
add-labelAdds a key=value label to a run.
list-step-templatesLists the available workflow step templates.
get-step-templateGets the details and input schema of one step template.
check-secretsCompares the environment variables of a collection with the variables a runbook needs.
set-environment-varsSets environment variables on a collection. Existing keys are kept. A null value deletes a key.
list-routinesLists the schedules of a collection, or of one task.
get-routineGets one schedule, including its next fire time.
create-routineCreates a schedule for an existing task.
update-routineChanges the cadence, overrides, secrets, paused state, or webhook of a schedule.
delete-routineDeletes a schedule. This cannot be undone.
pause-routineStops a schedule until it is resumed.
resume-routineResumes a paused schedule.
run-routine-nowFires a schedule immediately and returns the workflow_id of the run.
list-routine-runsLists the recent runs that a schedule started.
BEFORE YOU INSTALL
What to know first
Tools you get
27 tools named in the project's documentation.
Requires
- Node.js 18 or later
- a Jetty account and an API token from Settings > API keys (it starts with mlc_); Jetty's free plan includes 12 runs a month with no card, and paid plans start at $599 a month
INSTALLATION
Choose how to get started.
Direct Install
Claude Code
TERMINAL
claude mcp add jetty -- npx -y jetty-mcp-serverRun this in your terminal to add the server to Claude Code
MCP Clients
Only the clients this project documents.
Claude Code — plugin
CLAUDE CODE — PLUGIN · TERMINAL
claude plugin marketplace add jettyio/jettyio-skills
claude plugin install jetty@jettyRun both in a terminal, then type /jetty-setup in Claude Code once. This also brings Jetty's four skills, and the wizard saves the token the server reads.
Cursor, Windsurf
JSON
{
"mcpServers": {
"jetty": {
"command": "npx",
"args": [
"-y",
"jetty-mcp-server"
],
"env": {
"JETTY_API_TOKEN": "<YOUR_JETTY_API_TOKEN>"
}
}
}
}VS Code
VS CODE · JSON
{
"servers": {
"jetty": {
"command": "npx",
"args": [
"-y",
"jetty-mcp-server"
],
"env": {
"JETTY_API_TOKEN": "<YOUR_JETTY_API_TOKEN>"
}
}
}
}Codex
CODEX · TOML
[mcp_servers.jetty]
command = "npx"
args = ["-y", "jetty-mcp-server"]
env = { JETTY_API_TOKEN = "<YOUR_JETTY_API_TOKEN>" }Gemini CLI
GEMINI CLI · TERMINAL
gemini extensions install https://github.com/jettyio/jettyio-skills