Skip to main content
Glama

terminal

Execute terminal commands and wait for results. Use the run action to send input, wait for a pattern or quiet period, then return output. For incremental reading or sending, use read/send actions.

Instructions

Purpose: Interact with a terminal window: read output, send input, or run+wait+read in one call. action='read' / action='send' absorb the formerly-standalone read/send tools (Phase 4). Details: action='run' is the recommended high-level workflow: send command → wait until quiet/pattern/timeout → read output. The command text is passed as input (the legacy parameter name command is also accepted as a deprecated alias — see issue #245). Returns completion={reason, elapsedMs} first-class plus outputIntegrity:'ok'|'baseline_lost' so callers can detect when scrollback could not be anchored to the pre-send buffer. action='read' reads current text via UIA TextPattern (falls back to OCR); use sinceMarker for incremental diff. action='send' sends a command with focus management. Prefer: action='run' for command execution + result. For long-running commands (test runners, builds, deploys) use until:{mode:'pattern', pattern:''} — the default quiet mode is tuned for short interactive commands and may complete prematurely on multi-second silent gaps mid-run. Use action='read'/'send' for fine-grained control or when you need to interleave other actions. Caveats: Do not screenshot the terminal — action='read' is cheaper and structured. action='run' supports completion reasons: quiet | pattern_matched | exited | timeout | window_closed | window_not_found | send_failed (send rejected on a live window — see warnings for the underlying error code). until:{mode:'exit', shell:'bash'|'powershell'} (issue #386) returns completion.exitCode + reason:'exited' via an echo-immune sentinel that works for multiline input that pattern mode cannot anchor; pass shell explicitly (auto fails as ExitModeShellAmbiguous on WT/conhost/SSH), cmd is unsupported (ExitModeShellUnsupported), open-construct input is rejected (ExitModeUnsafeInput). When outputIntegrity:'baseline_lost' is returned, output is forced to '' and readError.code='BaselineMarkerLost' is set: rerun with until:{mode:'pattern',...} or longer timeoutMs. action='run' may also emit warnings prefixed FileLockCollision: when output reveals an EBUSY/Windows-lock/EAGAIN-EDEADLK file collision (e.g. shell '>' redirect colliding with the script's own writer — issue #236). Default quietMs=1500 (issue #196); long silences require pattern mode. preferClipboard=true (send default) overwrites clipboard. Hidden-input prompts emit verifyDelivery.unverifiable (reason:'hidden_input_prompt') — use method:'foreground'. action='read' typed errors: TerminalWindowNotFound, TerminalTextPatternUnavailable (force source:'ocr'); stale sinceMarker → hints.terminalMarker.previousMatched:false on ok:true (omit sinceMarker). FG-path Win11 foreground refusal returns code:'ForegroundRestricted' — switch to method:'background' or DTM_BG_AUTO=1. BG path auto-engages only when (a) the target window class is ConsoleWindowClass (conhost: cmd / PowerShell / pwsh classic hosts) OR (b) env DTM_BG_AUTO=1 is set globally. Windows Terminal (CASCADIA_HOSTING_WINDOW_CLASS) is intentionally EXCLUDED from auto-engage (issue #173): WT runs on WinUI/XAML and silently drops WM_CHAR posted to its HWND, so the FG path is used by default — pass sendOptions:{method:'background'} only if you have verified your WT build accepts BG input. Examples: terminal({action:'run', windowTitle:'PowerShell', input:'npm test', until:{mode:'pattern', pattern:'Test Files'}}) → recommended for test runners; matches when vitest summary appears terminal({action:'run', windowTitle:'pwsh', input:'ls'}) → quiet 1500ms wait, returns output (short interactive) terminal({action:'run', windowTitle:'pwsh', command:'ls'}) → identical to the above; command is a deprecated alias of input (issue #245) terminal({action:'read', windowTitle:'PowerShell', sinceMarker:'...'}) → incremental diff using the read action terminal({action:'send', windowTitle:'PowerShell', input:'echo hello'}) → sends text + Enter using the send action

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
actionYesAction selector — one of: read, send, run. Per-action required fields are enforced at call time (see the tool description); this flat schema lists every action's fields as optional.
includeNoOptional response-shape opt-in. `['envelope']` returns the self-documenting envelope (`_version` / `data` / `as_of` / `confidence`). `['raw']` forces raw shape (overrides DESKTOP_TOUCH_ENVELOPE=1 server default). Default behaviour is raw shape (compat with existing clients).
windowTitleNoPartial title of the terminal window (e.g. 'PowerShell', 'pwsh', 'WindowsTerminal').
inputNoCommand to send (Enter is appended automatically). Either `input` or its deprecated alias `command` is required.
commandNo[Deprecated alias of `input`] Accepted for callers that mis-remember the parameter name; new code should use `input`. If both are set, `input` wins.
untilNo
timeoutMsNoHard timeout in ms (default 30s)
sendOptionsNoExtra options forwarded to terminal send (method, chunkSize, etc.)
readOptionsNoExtra options forwarded to terminal read (lines, source, ocrLanguage, etc.)
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations, the description carries full burden. It comprehensively discloses behavior: completion reasons, output integrity, FileLockCollision warnings, hidden-input prompts, foreground restrictions, background auto-engage conditions, and platform-specific behavior (conhost vs Windows Terminal). No contradictions.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Well-structured with sections (Details, Prefer, Caveats, Examples) but verbose. Every sentence adds value given tool complexity, but could be more condensed. Still, front-loaded with purpose and recommendations.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Covers all actions, parameters, return behavior, warnings, edge cases, and examples. Despite no output schema, explains output integrity and completion details. Thorough and complete for a complex tool with 9 parameters and nested objects.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters5/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 89%, but the description adds significant value beyond schema descriptions: explains `input` usage, deprecated `command`, `until` modes in detail, `include` options, `windowTitle` matching, and `readOptions`/`sendOptions` context. Provides examples and caveats.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states the tool's purpose: 'Interact with a terminal window: read output, send input, or run+wait+read in one call.' It distinguishes three actions (read, send, run) and explains the recommended workflow. This differentiates it from sibling tools like browser or clipboard.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Provides extensive guidance: 'Prefer: action='run' for command execution + result.' For long-running commands, recommends using until with pattern mode. Warns against screenshotting the terminal. Details when to use read/send for fine-grained control. Explains foreground vs background methods and when each applies.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Install Server

Other Tools

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/Harusame64/desktop-touch-mcp'

If you have feedback or need assistance with the MCP directory API, please join our Discord server