Core vs Glasswarp
A side-by-side look at the Core and Glasswarp MCP servers: transport, auth, tools, quality signals, and the config to install either one in Claude, Cursor, Windsurf or VS Code.
Verdict: choose Core for os automation work as a local process; choose Glasswarp for os automation as a local process. You can also run both side by side in one mcpServers block.
Which MCP server should you choose?
Choose Core when
- You need dedicated capabilities in the OS Automation domain.
- You prefer a local stdio subprocess transport.
- You want tools like
electron_assert_pattern, electron_attach, electron_check.
Choose Glasswarp when
- You need dedicated capabilities in the OS Automation domain.
- You prefer a local stdio subprocess transport.
- You have the keys it needs:
GLASSWARP_API_KEY. - You want tools like
demos.list, demos.get, rigs.list.
Feature & specification comparison
Rows where the two servers differ are marked.
| Spec | ||
|---|---|---|
| Category | OS Automation | OS Automation |
| Quality signal | 57/100 · Good | 57/100 · Good |
| Install command | Confirmed | Confirmed |
| Transport | Local subprocess (stdio) | Local subprocess (stdio) |
| Auth | Not declared | Not declared |
| Pricing | Not declared | Not declared |
| Required env vars | None listed | GLASSWARP_API_KEY |
| Verified owner | Community listing | Community listing |
| GitHub stars | 12 | 0 |
| Directory activity | 1 views 0 installs 0 upvotes | 2 views 0 installs 0 upvotes |
| Tools | 55 | 16 |
| Full listing | View Core | View Glasswarp |
Tools & capabilities
Core tools 55
electron_assert_patternValidate, in a single check (no polling), that an element's text or a named attribute matches a pattern. Target by ref or selector. With `attribute` set, reads that attribute; otherwise reads the element's trimmed text. Provide exactly one of: equals, contains, matches_regex. Optional flags (any of i, m, s, u) apply to matches_regex; g and y are rejected as stateful. Returns: { ok, session_id, matched, actual }. Errors: EXPECTATION_FAILED (element found but its value did not match the pattern — details carry expected + actual; a missing attribute reads as actual: null), SELECTOR_NO_MATCH (no element matched — this is one-shot, so a missing element is a precondition failure, not a retry; carries similar_refs), REF_NOT_FOUND (stale ref), TRANSPORT_UNSUPPORTED, NOT_RUNNING, BAD_ARGUMENT (no/multiple predicates, invalid regex or flags, or ref+selector both/neither).electron_attachAttach to an already-running Electron app exposing a CDP debug endpoint (use electron_discover_running to find one, or start the app with --remote-debugging-port). Provide port (+ optional loopback host) or a loopback cdpUrl; pid alone is not attachable over CDP but, when supplied alongside, lets stop escalate to SIGKILL. The CDP transport supports eval/read/observe and core interaction surfaces against the attached app. The selected CDP page is also the default snapshot/find surface; explicit iframe/webview surface targeting remains unavailable. Returns: { ok, session_id, transport, windows, capabilities }. Errors: TRANSPORT_UNSUPPORTED (no attach-capable transport), CDP_DISCONNECTED (endpoint unreachable or dropped; retryable), CDP_TIMEOUT (handshake/method timeout; retryable), BAD_ARGUMENT (missing target selector or non-loopback endpoint).electron_checkCheck the checkbox/radio identified by ref or selector (no-op if already checked). Options: force, timeoutMs. Returns: { ok, session_id, target, checked: true }. Errors: REF_NOT_FOUND / SELECTOR_NO_MATCH (carries similar_refs), ELEMENT_NOT_VISIBLE (retryable), ELEMENT_DISABLED, NOT_RUNNING, BAD_ARGUMENT.electron_clear_inputClear the value of the input/textarea identified by ref or selector (sets it to empty). Options: force, timeoutMs. Returns: { ok, session_id, target, cleared }. Errors: REF_NOT_FOUND / SELECTOR_NO_MATCH (carries similar_refs), ELEMENT_NOT_VISIBLE (retryable), ELEMENT_DISABLED, NOT_RUNNING, BAD_ARGUMENT.electron_clickClick the element identified by ref (from a snapshot) or selector. Options: button (left|right|middle, default left), clickCount (2 = double-click), force (bypass actionability), timeoutMs. Returns: { ok, session_id, target }. Errors: REF_NOT_FOUND / SELECTOR_NO_MATCH (no such element — re-snapshot; not retryable, carries similar_refs), ELEMENT_NOT_VISIBLE (retryable), ELEMENT_DISABLED (not retryable), NOT_RUNNING, BAD_ARGUMENT (ref+selector both/neither).electron_console_logsRead the captured renderer console output for the session, newest-relevant entries last. Filters (all optional, ANDed): type (one or more of log/info/warning/error/debug/...), match (a regular expression the text must match), since (epoch ms — only entries at/after it), limit (max entries, default 200, max 1000 — the most recent are kept). Returns: { ok, session_id, entries: [{ type, text, timestamp, windowId?, location? }], count, overflowed }. overflowed is the number of older entries the buffer dropped. Errors: NOT_RUNNING, BAD_ARGUMENT (invalid regex, or multiple sessions).electron_detachDisconnect from an attached or injected app without stopping it. A launch-owned session cannot detach because its transport owns the process; this includes Playwright development launches and packaged CDP launches. Use electron_stop for that case. Returns: { ok, session_id, detached: true }. Errors: TRANSPORT_UNSUPPORTED (launch-owned session; not retryable), NOT_RUNNING (no such session), BAD_ARGUMENT (multiple sessions).electron_dialog_handlerArm the auto-responder for native JS dialogs (alert/confirm/prompt/beforeunload) and read which dialogs fired. Dialogs block the renderer, so the policy is applied automatically the instant one appears. Arming args (all optional): action (accept|dismiss — the default for every dialog), perType (per-kind overrides, e.g. {"confirm":"accept","beforeunload":"dismiss"}, falls back to action), promptText (text submitted to prompt() when it is accepted), oneShot (apply to exactly the next dialog, then revert to dismiss). With NO arming args the call is inspect-only and leaves the policy unchanged. Read args (all optional): type (one or more kinds to include), since (epoch ms), limit (max events, default 50, max 200 — most recent kept), clear (flush the whole buffer after reading). Until armed, the default policy is dismiss, so dialogs never hang the app. Returns: { ok, session_id, policy, entries: [{ type, message, action, defaultValue?, promptText?, timestamp, windowId? }], count, overflowed }. overflowed counts dropped events across the whole buffer, not just the returned (type/since/limit-filtered) subset. Errors: NOT_RUNNING, BAD_ARGUMENT (promptText without an accepting prompt policy, or oneShot without a policy to arm).electron_discover_runningScan the conventional CDP debug ports (9222-9225 by default) for already-running, debuggable Electron apps on loopback only. No session required. Returns: { ok, targets, count, scanned } where each target is { targetId, port, appName, pid } and scanned reports { host, ports, elapsed_ms } so an empty result is unambiguous. Errors: BAD_ARGUMENT (non-loopback host, invalid port list, or timeout outside bounds). A failed probe is simply "no target on that port".electron_doctorRun non-mutating preflight checks without starting an Electron session: Node version, Playwright, Electron, Linux display, configured app root and screenshot directory, eval policy, and project Electron runtime alignment. Returns: { ok, doctor_ok, checks, runtime }, where runtime includes server and, with --app-root, target Electron/Node/V8/ABI facts plus a bounded native-addon inventory. Inspect failed checks and warnings before electron_launch. Errors: none.electron_dragDrag the source element (ref or selector) onto the target element (targetRef or targetSelector), using the real mouse API. Options: force, timeoutMs. Returns: { ok, session_id, source, target }. Errors: SELECTOR_NO_MATCH / REF_NOT_FOUND (carries similar_refs), ELEMENT_NOT_VISIBLE (retryable), NOT_RUNNING, BAD_ARGUMENT (a side missing or both ref+selector given).electron_drop_fileSimulate dropping OS files onto the element identified by ref or selector. Web DataTransfer mode: reads each path on the host running the server, rebuilds the files in the renderer, and dispatches dragenter/dragover/drop with a real DataTransfer — engaging standard web drop handlers. Paths must be ABSOLUTE (max 10 files, 5242880 bytes each). default_prevented reports whether a drop handler engaged (called preventDefault); false usually means the target has no web drop handler. Apps that resolve dropped files to native OS paths need an app-specific IPC convention this tool does not simulate. Options: mimeType, timeoutMs. Returns: { ok, session_id, target, files, default_prevented }. Errors: ABSOLUTE_PATH_REQUIRED, FILE_NOT_FOUND, BAD_ARGUMENT (too many/large files, or ref+selector both), SELECTOR_NO_MATCH / REF_NOT_FOUND (carries similar_refs), NOT_RUNNING.- +43 more on the listing page
Glasswarp tools 16
demos.listList showcase run contracts (id, title, install, command). Read-only catalog — does NOT start a session, touch a rig, or run solvers. Use when the user asks for Minesweeper/Mona Lisa/Paint demos or you need the glasswarp-demo command. Prefer demos.get for one full card. For ad-hoc UI work use rigs.list → session.start → screen.observe instead.demos.getReturn one showcase run contract (install, command, needs, framing). Does NOT execute the demo or control a PC. Call after demos.list when you know the demo_id. If the client can run shell, offer the command; if chat-only, show the card. Do not replace this with a slow MCP click loop for solver demos.rigs.listList Windows machines (rigs) paired to this API key: id, name, online, api_access_enabled, and USABLE flag. Read-only — does not start a session. Call first before start_session. A rig is USABLE only when online AND the owner enabled API access. If none are USABLE, tell the user to install the host agent, pair in Console → Rigs, and enable API access — never ask for OS passwords.session.startStart a metered desktop session on a USABLE rig from rigs.list. Side effects: begins wall-clock billing, shows an on-screen “API session active” indicator, enables observe/input until session.end. Idle sessions auto-end after ~15 minutes. Always call session.end when done or abandoning. Do not call if no USABLE rig exists. Returns session_id and Live View URL (owner console login required).session.endEnd an active session. Side effects: stops billing, runs host safety_restore, closes apps launched via app.launch. Always call when finished or abandoning — do not leave sessions open. Safe to call once; further observe/input on that session_id will fail.screen.observeRead the current screen: UIA targets (numbered ids + native coords) and a text summary. Does not move mouse/keyboard. Default image=false (no JPEG) for speed; set image=true only when you must judge pixels visually (then max_width≈960, quality≈60). If changed=false, JPEG is omitted even when requested — do not re-analyze; wait or act differently. If dirty is null, assume changed. Prefer input.send_actions for multi-step UI; observe after meaningful steps, not after every click. Target ids are valid only until the next UI change.input.click_targetLeft/right/middle-click a UIA target by id from the latest screen.observe (uses native center coords). Prefer over input.click_xy. Side effect: real mouse click on the remote Windows desktop. Do not reuse target_id after the screen may have changed — re-observe first. For click→type→keys sequences, use input.send_actions (one turn) instead of chaining this tool.input.click_xyClick at native screen coordinates (0…native_width-1, 0…native_height-1 from screen.observe). Last resort when no suitable UIA target exists — prefer input.click_target. Never use JPEG/downscaled pixel coords. Side effect: real mouse click on the remote desktop.input.type_textType a Unicode string into the currently focused control via native input. Does not click first — focus the field (input.click_target / input.send_actions) before calling. Side effect: keystrokes on the remote desktop. For form fills (click → type → tab/enter), prefer input.send_actions in one call.input.send_keysSend a key or chord to the focused window (e.g. enter, tab, ctrl+s, alt+f4, win). Side effect: real key events on the remote desktop. Prefer bundling into input.send_actions when the shortcut follows a click/type in the same planned sequence. Use input.type_text for literal strings, not this tool.input.dragPress-move-release mouse drag in native capture coordinates. Use for drawing, sliders, selection boxes, and drag-and-drop. Side effect: mouse_down → moves → mouse_up on the remote desktop. Prefer input.send_actions if the drag is one step in a longer predictable sequence.input.scrollMove the cursor to native (x,y) then apply a vertical mouse-wheel delta. Side effect: scroll on whatever is under that point. Negative delta scrolls toward the bottom of the page. Prefer input.send_actions when scroll is part of a multi-step sequence. Re-observe after scrolling lists/pages before clicking targets.- +4 more on the listing page
Ready-to-paste client configs
Paste either block (or both) into your client config, e.g. claude_desktop_config.json or ~/.cursor/mcp.json. We only show a config when the install command is confirmed.
Core
{
"mcpServers": {
"core": {
"command": "npx",
"args": [
"-y",
"--package",
"@electron-stagewright/core@0.5.0",
"--package",
"playwright@1.61.1",
"--package",
"electron@42.3.0",
"electron-stagewright"
]
}
}
}
Glasswarp
{
"mcpServers": {
"glasswarp": {
"command": "npx",
"args": [
"-y",
"@glasswarp/mcp"
],
"env": {
"GLASSWARP_API_KEY": "YOUR_GLASSWARP_API_KEY_HERE"
}
}
}
}
Frequently asked questions
Core is categorized under OS Automation and uses a local stdio subprocess. Glasswarp belongs to OS Automation and uses a local stdio subprocess. Pick Core for os automation work and Glasswarp for os automation.
Copy the mcpServers block from the config section on this page into your client's config (for example claude_desktop_config.json or ~/.cursor/mcp.json), then fully restart the client.
Core has no declared pricing — check the project's own site. Glasswarp has no declared pricing — check the project's own site. If a server needs environment variables such as API keys, define them under the env key in your client's MCP config.
Yes. MCP clients support multiple servers: add core and glasswarp as separate keys in the same mcpServers object, and your assistant will call tools from either one.
On AllMCPs, Core has a quality signal of 57/100 (Good) with 1 views, 0 installs and 12 GitHub stars. Glasswarp has 57/100 (Good) with 2 views, 0 installs and 0 GitHub stars.