JM Tech Power

jm-tech-power-official/jm

jm — the JM Tech Power command-line client. Manage your apps, releases, JMDrive files and chat from the terminal, like git/gh. Pure Python 3, no dependencies. Install, then run 'jm auth login --web' (Google/GitHub friendly).

Public Developer Tools Linux, macOS, Windows MIT
README.md

jm — JM Tech Power CLI

Control your JM Tech Power account from the terminal, like git / gh. Pure Python 3.7+, no dependencies.

Install

Requires Python 3 (the installers check for it and tell you how to get it).

Linux / macOS — one line (installs to /usr/local/bin, cleans up after itself):

curl -fsSL https://www.jmtechpower.com/jm/install.sh | sudo bash

Windows — in PowerShell:

irm https://www.jmtechpower.com/jm/install.ps1 | iex

Or from the downloaded zip: sudo ./install.sh (Linux/macOS) or .\install.ps1 (Windows).

Watch what happens (events + webhooks)

jm events # what happened since you last looked jm events --follow # keep watching jm events -f --json # one JSON object per line, for piping jm events -f --exec ./deploy.sh # run something for each event

Following is cursor-based, so nothing is missed or replayed between polls.

Projects can also push to you:

jm webhook add owner/name https://example.com/hook jm webhook add owner/name https://e.com/h --events release_published,issue_opened jm webhook list owner/name

Each delivery carries X-JM-Signature: v1=HMAC_SHA256(secret, timestamp + "." + body) and X-JM-Timestamp. Verify both, and reject timestamps far from now — that is what makes a replayed delivery detectable. Failed deliveries retry with exponential backoff, and an endpoint that keeps failing is disabled rather than retried forever.

Run someone else's project in one command

jm run owner/name

Downloads the release, reads its jm.toml, and starts it. Nothing runs until you have seen the exact commands and said yes — and if a project has no jm.toml, jm stops rather than guessing at a start command.

Publishing your own? Scaffold the manifest:

jm run --init

[run] command = "python3 main.py" install = "pip install -r requirements.txt" venv = true # install and run inside a private virtualenv

Use it from an AI agent (MCP)

jm is also a Model Context Protocol server, so Claude, Cursor and other agents can work against your account directly — search projects, read source, open issues, publish releases.

jm mcp install # prints the config snippet for your client

For Claude Code, one line:

claude mcp add jm -- jm mcp serve

It authenticates with the same login you already did, so there is no second credential to manage. Tools that write (open_issue, create_release) say so in their description, so a well-behaved agent will check with you first.

Log in

jm auth login # opens your browser — click "Authorize" and you're in (like gh)

Works with any sign-in method, including Google / GitHub (no password needed). Alternatives:

jm auth login --with-token # paste a token from https://www.jmtechpower.com/settings/tokens jm auth login --password # username + password (accounts that have one)

Config lives in ~/.config/jm/config.json. Env overrides: JM_HOST, JM_TOKEN.

Commands

jm whoami jm account [--nickname N] [--bio B] # show / update account settings jm readme | help # show this document jm update [--check] # update jm to the latest version

jm search <query> # JM Search: projects, articles & people

# Projects — one namespace for the whole lifecycle (GitHub-style) jm proj list [--mine] [--search QUERY] # browse projects jm proj search <name> # find projects by name jm proj install <owner/slug | name | id> [-y] [--dir DIR] [--download-only] jm proj installed # projects you've installed jm proj update [-y] [--check] # upgrade installed projects jm proj uninstall <project> # remove + untrack

# Version control — native to a project (a local .jm object store) jm proj init # start a repository here jm proj add <path>... | rm <path> | mv <src> <dst> jm proj commit -m MSG [-a] | status | show [<ref>] | log [-n N] jm proj diff [<a>] [<b>] [--stat|--name-only] | blame <path> jm proj branch [NAME | -d NAME] | checkout <ref> | tag [NAME | -d NAME] jm proj merge <branch> | rebase <upstream> [--continue|--abort] jm proj cherry-pick <ref> | revert <ref> | reset [--soft|--mixed|--hard] <ref> jm proj stash [pop | apply | list | drop] jm proj remote [set <owner/slug>] # where push sends commits jm proj clone <project> [--dir D] [--version V] # get a project's code jm proj push [--branch B] | fetch | pull [--branch B]

# Release source — a browsable source ZIP per version (published on the website) jm proj release new <project> --version V [--notes N] [--file F] [--asset PATH[:PLATFORM/ARCH]]... # publish a release jm proj release clone <project> [--dir D] [--version V] # source ZIP, not the repo jm proj release ls <project> [--version V] | cat <project> <path> [--version V] jm proj release releases <project> | view <project> | readme <project> jm proj release log <project> | diff <project> [v1] [v2] [--stat|--name-only] jm proj release checkout <project> <version> [--dir D] | pull [--dir D]

jm proj visibility <project> <public|private> # change who can see it jm proj delete <project> # delete a project you own

# Social (also usable at top level: jm star <project>, jm issues <project>, …) jm proj star <project> | unstar <project> jm proj watch <project> [--level releases|all] | unwatch <project> jm proj issues <project> [--state open|closed|all] [--json] jm proj issue <project> <number> [--json] jm proj issue-create <project> --title T [--body BODY|-] jm proj issue-comment <project> <number> [--body BODY|-] jm proj issue-close <project> <number> | issue-reopen <project> <number> jm notifications [--unread] [--json] (alias: jm inbox) jm notifications read [--all] # mark notifications read

jm articles list [--mine] [--search QUERY] [--category C] jm article view <id> [--raw] jm article create --title T --category C (--content "..." | --file body.html | stdin) [--summary ...] [--tags a,b] [--draft] jm article edit <id> [--title ...] [--content ... | --file ...] [--tags ...] [--publish | --unpublish] jm article delete <id>

jm drive ls [folder_id] jm drive upload PATH... [--folder ID] [--public] jm drive mkdir NAME [--parent ID] [--public] jm drive download <id> [-o out] # a folder downloads as a zip jm drive share <id> [--password P] [--limit N] jm drive rm <id>

jm chat groups jm chat read (--with USER_ID | --group ID) [--limit N] jm chat send "message" (--to USERNAME | --group ID) jm chat retract <message_id> # unsend one of your own messages

jm token list | create [--name N] | revoke <id>

jm link [--name N] # run as a device for website "Direct Install" jm devices | unlink <id> # list / remove linked machines sudo jm service install | uninstall # run jm link at boot (recommended) jm service status # no sudo needed

jm admin stats | apps | users # official account only jm admin approve <id> | revoke <id> | delete <id>

jm star <project> | unstar <project> # star / unstar a project jm watch <project> [--level releases|all] | unwatch <project> jm issues <project> [--state open|closed|all] [--json] jm issue <project> <number> [--json] jm issue-create <project> --title T [--body BODY|-] jm issue-comment <project> <number> [--body BODY|-] jm issue-close <project> <number> | issue-reopen <project> <number> jm notifications [--unread] [--json] (alias: jm inbox) jm notifications read [--all] # mark notifications read

Projects

jm proj is the one command for everything project-related — version control, publishing, installing, and the social layer — from the terminal. jm project is a long alias for jm proj.

The top-level verbs are version control (jm proj init/commit/branch/merge/ push/pull/…, backed by a local .jm object store). Release source lives under its own group, jm proj release … (a browsable source ZIP per version — clone it, browse it, diff it). A project has both: real commit history and releases, plus stars, watchers, forks, pull requests and issues. Creating a project and publishing a release are done on the website.

Track your work locally with version control:

jm proj init # start a repository jm proj add . && jm proj commit -m "first cut" jm proj branch feature && jm proj checkout feature jm proj merge feature # line-level 3-way merge

Remotes

A repository is named <owner>/<name>.jmproj, the way GitHub names one <owner>/<name>.git. Point a working copy at one and push:

jm proj remote set jm-tech-power-official/quill.jmproj jm proj remote # show the current remote jm proj push # no argument needed once it is set

Every way of writing it resolves to the same project, so paste whichever you have to hand:

jm proj push jm-tech-power-official/quill jm proj push jm-tech-power-official/quill.jmproj jm proj push https://www.jmtechpower.com/jm-tech-power-official/quill.jmproj

jm proj clone takes the same forms and records the name for you:

jm proj clone jm-tech-power-official/quill.jmproj --dir ./quill

Opening a .jmproj URL in a browser lands on the project page, so the string you copy works whether it ends up in a terminal or an address bar. The project's Source tab shows what you pushed — the file tree at the tip of a branch, with a branch and tag switcher — rather than the contents of a release ZIP.

Working copies made before remotes had names store a numeric project id. They keep working; jm proj remote will show the id and tell you how to name it.

Projects are created and releases are published on the website (https://jmtechpower.com) — you create a project there (metadata only, no upload required), then use the project page to upload its browsable source files and to draft a release with installers and notes. The CLI's job is the other half of the loop: clone a project and work with it locally, browse or diff any published version, and install releases.

jm proj clone jm-tech-power-official/quill # get the code locally jm proj pull # fetch + merge new commits jm proj release releases quill # see what's been published

jm proj clone is the one to reach for. It clones the repository when the project has one, and falls back to the latest release source when nobody has pushed commits — so you do not have to know which kind of project it is before you type the command. --version V always means the published source at that version.

Two files describe a local working copy:

  • jm.toml — the project manifest. A tiny flat TOML of key = "value"

lines: name, version, category, supported_os, license, description, plus id, owner and slug. It is written for you by jm proj release clone / checkout (and refreshed by pull); there is no init command that scaffolds it. jm proj release pull needs the id.

  • .jmignore — gitignore-lite patterns for files to keep out of the

repository (honored by jm proj add, status, commit and friends). Each non-empty, non-# line is an fnmatch pattern; a trailing / means a directory. When the file is absent these defaults apply: .git/, __pycache__/, .pyc, .DS_Store, node_modules/, .venv/, venv/, dist/, build/, .zip.

jm proj release clone is the explicit way to ask for a published source zip rather than the repository — useful when a project has both and you want a specific version. It unpacks the zip and writes a jm.toml. Accepts a project as an id, a name, or owner/slug:

jm proj release clone jm-tech-power-official/quill jm proj release clone 42 --dir ./mytool --version 1.2.0

Browse a published project's source without cloning:

jm proj release ls my-tool # one file path per line jm proj release cat my-tool src/main.py # print a file to stdout jm proj release releases my-tool # versions, sizes, downloads, source? jm proj release view my-tool # a one-screen summary

All read commands (ls, cat, releases, view) also take --json. Every project argument resolves as an id, a name, or owner/slug (owner/slug is looked up via the projects API).

Install a project's latest release for your OS, or change its visibility:

jm proj install jm-tech-power-official/quill # by owner/slug jm proj install 30 --download-only # just download project #30 jm proj visibility my-tool private # make it private jm proj visibility my-tool public # …or public again

Social

Star projects, watch them for updates, file issues and read your inbox — GitHub-style, from the terminal. Everything lives under jm proj (and the same verbs also work at the top level: jm star, jm issues, …). Every command takes a project as an id, a name, or owner/slug.

jm proj star my-tool # star a project (prints the new star count) jm proj unstar my-tool # remove your star

jm proj watch my-tool # get notified about new releases jm proj watch my-tool --level all # …about releases, issues and comments jm proj unwatch my-tool # stop watching

Issues:

jm proj issues my-tool # open issues (a table) jm proj issues my-tool --state all --json # every issue, as JSON jm proj issue my-tool 7 # view issue #7 with its comments jm proj issue-create my-tool --title "Bug: crash on start" --body "Steps: …" echo "long body" | jm proj issue-create my-tool --title "…" --body - # body from stdin jm proj issue-comment my-tool 7 --body "I can repro this." jm proj issue-comment my-tool 7 --body - # comment body from stdin jm proj issue-close my-tool 7 jm proj issue-reopen my-tool 7

--body - (or piping on stdin) reads the body from standard input, so you can compose long text in a file: jm proj issue-create my-tool --title Foo --body - < body.md.

Notifications are your event bus — agents can poll them as clean, stable JSON:

jm notifications # recent notifications (● = unread) jm inbox # alias for jm notifications jm notifications --unread # only the unread ones jm notifications --unread --json # machine-readable; stable for scripting/agents jm notifications read # mark all as read jm notifications read --all # (same) mark everything read

Examples

jm proj clone jm-tech-power-official/quill # clone a project's code jm proj release diff quill 1.0.0 1.1.0 --stat # what changed between versions

echo "<p>Hello world</p>" | jm article create --title "Hi" --category News --draft

jm drive upload report.pdf --public # prints a https://.../d/<token> link

jm proj install "Quickfire" # download the project and run its installer jm proj install 12 --download-only # just download project #12

Installing projects

jm proj install <owner/slug | name | id> downloads a project's latest release and installs it for your OS: .deb/.rpm via the package manager, .AppImage into ~/.local/bin, .pkg/.dmg on macOS, .exe/.msi on Windows; archives (.zip/.tar.gz) are extracted. It asks before running an installer (skip with -y); --download-only just saves the file.

Projects that ship for several platforms

Attach one build per machine to the release and jm proj install picks the right one — a Linux box is never handed the .msi.

From the terminal, publish the whole release in one command:

jm proj release new myname/myapp --version 1.2.0 \ --notes "Faster startup" \ --asset dist/myapp-linux-x86_64.tar.gz \ --asset dist/myapp-linux-arm64.tar.gz \ --asset dist/MyApp-arm64.dmg \ --asset dist/MyApp-Setup.exe \ --asset dist/SHA256SUMS.txt

--asset repeats, once per file. Each file's target is read from its name; add one explicitly when the name does not say, with PATH:PLATFORM/ARCH:

--asset dist/build.zip:linux/arm64

Platforms are linux, macos, windows, android, ios; architectures are x86_64, arm64, x86. Either half may be left out — :macos/ is any Mac.

--file is the single main installer, for projects that have one. You do not need it: a release made of per-platform builds alone is complete, and anyone on a machine you did not build for falls back to --file if there is one.

On the website the same thing is the Draft a new release form: Main installer for the one file, Additional files for the per-platform builds — you can select several at once. Each is labelled from its name, and the project page shows the label beside every file with a control to correct it.

Each file's target is read from its name when it is uploaded, which is where projects already put it:

myapp-linux-x86_64.tar.gz -> Linux x86-64 myapp-linux-arm64.tar.gz -> Linux arm64 MyApp-arm64.dmg -> macOS arm64 MyApp-Setup.exe -> Windows SHA256SUMS.txt -> no platform, so never installed

A file with nothing to go on — a plain build.zip — is left unassigned, and the project owner sets it on the project page, next to the file. The same controls fix a wrong guess.

An exact platform-and-architecture build is preferred, then one that names only the platform. If nothing on the release fits your machine, jm falls back to the main installer, so a project that ships a single file keeps working exactly as it did.

If the developer marked an installer inside the package (or you pass --run), jm runs it after extracting:

jm proj install MyApp --run setup.exe # run setup.exe from inside the zip

Installed projects are remembered on your account (synced with the website, where the same projects show "✓ Installed" / update badges). Manage them:

jm proj update # install newer versions of everything you installed jm proj update --check # just list what has updates jm proj installed # show what's tracked, and on which machine jm proj uninstall <project> # remove it from disk (apt/dnf remove, delete files + install path) + untrack jm proj uninstall <project> --keep-files # just untrack, leave files

Developers can declare an uninstaller; jm proj uninstall runs it (then also removes the package + install path). An install location can be set two ways:

  • the developer declares it on the project (the "Install path" field on the

project's create/edit form on the website; the API also accepts install_path on POST /api/v1/apps)

  • the installer/user reports the chosen path at install time:

jm proj install <project> --install-path "/opt/myapp" or POST it to the API: POST /api/v1/installed {app_id, version, path, device} jm proj uninstall then deletes that path too (with a confirmation + a safety guard).

Devices and Direct Install

Link a machine to your account and the website can install projects onto it: open a project page, press Direct install, pick the machine, and watch it go. The machine makes the connection outbound, so this works from behind NAT with no ports opened.

sudo jm service install # run the agent at boot, before anyone logs in jm service status # no sudo needed sudo jm service uninstall

This installs a system service — a unit in /etc/systemd/system on Linux, a LaunchDaemon on macOS, or an at-startup Scheduled Task on Windows — so the machine is reachable from the website after a reboot without anyone signing in. Installing one needs root; checking its status does not.

It runs as you, not as root. The website can queue install jobs onto a linked machine, and a root daemon taking orders from a web button is a much bigger thing to agree to than a user one. The practical consequence is unchanged from before: user-local installs work, and a job that needs apt or dnf still comes back as needs_admin for you to finish on that machine.

Run jm auth login on that machine first, as the user the service will run as — the agent reads its token from that account's ~/.config/jm.

jm link # same agent, in the foreground; stops when you close the window jm devices # the machines you have linked jm unlink <id> # remove one

jm link is the one to try first, but the service is what you want day to day: a terminal agent is offline exactly when you reach for the button.

A linked machine gets its own token, issued when it registers and scoped to that machine's install jobs — it cannot read your Drive, publish projects or do anything else on your account. Unlinking revokes it, from either side. A machine linked by a jm older than 2.7.0 still uses your account token until you re-link it; the Devices page on the website flags those.

Installs are recorded per machine, so the same project installed on a laptop and a desktop is two entries rather than one that overwrites the other:

jm proj installed # ID, NAME, INSTALLED, DEVICE, STATUS

The website's Installed apps page groups the same information by machine. Anything downloaded through the website, rather than installed by jm, is listed under Other — nothing told us which machine it landed on.

Deprecated: jm apps

jm apps ... still works as a thin alias for the matching jm proj ... commands, but it is deprecated and prints a one-line notice. Use jm proj.

Progress

Downloads and clones show a progress bar: jm proj install, jm proj clone, jm proj push and jm proj fetch. It is drawn on stderr and only when stderr is a terminal, so piping stdout into a file or another command still gets clean output and a log does not fill with redraws. Set JM_NO_PROGRESS=1 to turn it off anyway.

Updating

jm update # download + replace the installed binary in place (jm itself) jm update --check # just report whether a newer version exists

jm also checks once a day and prints a one-line notice when an update is available (disable with the JM_NO_UPDATE_CHECK env var). If jm lives in a system folder, run sudo jm update (Linux/macOS).