diff --git a/community/voice-dev-terminal/README.md b/community/voice-dev-terminal/README.md
new file mode 100644
index 00000000..4bf84c0c
--- /dev/null
+++ b/community/voice-dev-terminal/README.md
@@ -0,0 +1,102 @@
+# Voice Dev Terminal
+
+![Community](https://img.shields.io/badge/OpenHome-Community-orange?style=flat-square)
+![Author](https://img.shields.io/badge/Author-@hasni1731996-lightgrey?style=flat-square)
+
+A voice-native developer terminal for OpenHome. Say "terminal", then keep talking — run git commands, check disk space, find files, tail logs — all hands-free, with LLM-generated commands and spoken summaries of the output.
+
+## What Makes It Different
+
+| | Alexa / Siri | `openhome-local-link` template | Voice Dev Terminal |
+|---|---|---|---|
+| Session | One command, done | One command, done | Stays open for multiple commands |
+| Command source | Fixed set | LLM one-shot | LLM with full conversation history |
+| Output | Raw or none | Raw string spoken back | LLM-summarized into natural speech |
+| Dangerous commands | No guard | No guard | Voice confirmation required |
+| Inline trigger | No | No | "terminal, git status" runs immediately |
+
+## Trigger Phrases
+
+- `terminal` / `open terminal` / `dev terminal` / `developer terminal`
+- `run command` / `run a command` / `execute command`
+
+## Example Conversations
+
+**Basic session:**
+> "terminal"
+> → "Terminal ready. What do you need?"
+> "what's my git status?"
+> → "Running that now. You're on branch main, 2 files modified and not staged."
+> "show recent commits"
+> → "Running that now. 3 commits in the last day. Latest: fix auth bug, 2 hours ago."
+> "done"
+> → "Got it."
+
+**Inline trigger (no prompt needed):**
+> "terminal, check disk space"
+> → "Running that now. Your disk is 78% full — 56 gigs free of 256."
+
+**Dangerous command:**
+> "remove the build folder"
+> → "That runs: rm -rf ./build. Confirm?"
+> "yes"
+> → "Running that now. Done."
+
+**Context-aware follow-up:**
+> "list python files"
+> → "Running that now. Found 12 Python files in this directory."
+> "do it again but only in the src folder"
+> → "Running that now." (LLM resolves to `find ./src -name "*.py"`)
+
+## Exit Phrases
+
+`done` · `bye` · `goodbye` · `that's all` · `all done` · `close terminal` · `never mind`
+
+## Safety
+
+Dangerous command tokens (`rm`, `rmdir`, `sudo`, `shutdown`, `reboot`, `kill`, `pkill`, `dd`, `mkfs`) trigger a spoken confirmation before the command runs. Say anything other than "yes" to cancel.
+
+Commands run in your shell with your own permissions — the Local Link bridge provides no sandbox.
+
+## Setup
+
+This ability requires the **OpenHome CLI Local Link bridge** running on your machine.
+
+### 1. Install the CLI
+
+From the root of this repository (Python 3.10+):
+
+```bash
+python3 -m venv cli/.venv && source cli/.venv/bin/activate
+pip install -e cli
+```
+
+### 2. Log in
+
+```bash
+openhome login   # paste your API key from Dashboard → Settings → API Keys
+```
+
+### 3. Start the bridge
+
+```bash
+openhome local start    # runs in the background
+openhome local status   # confirm it's running
+openhome local logs     # watch commands and output in real time
+```
+
+Keep it running while you use the ability.
+
+### 4. Add the ability
+
+Add **Voice Dev Terminal** to your Agent from the Dashboard and set the trigger words above.
+
+## Notes
+
+- The LLM targets macOS zsh/bash. On Linux, most commands work as-is; adjust the system prompt in `get_system_prompt()` if needed.
+- Long-running commands time out after 30 seconds. Raise the timeout in `exec_local_command()` for things like `npm install` or large `find` operations.
+- Conversation history is in-memory and resets when the session exits. There is no persistent command log.
+
+## Author
+
+Muhammad Hassan
diff --git a/community/voice-dev-terminal/__init__.py b/community/voice-dev-terminal/__init__.py
new file mode 100644
index 00000000..e69de29b
diff --git a/community/voice-dev-terminal/main.py b/community/voice-dev-terminal/main.py
new file mode 100644
index 00000000..5f6d45f4
--- /dev/null
+++ b/community/voice-dev-terminal/main.py
@@ -0,0 +1,149 @@
+from src.agent.capability import MatchingCapability
+from src.agent.capability_worker import CapabilityWorker
+from src.main import AgentWorker
+
+HOTWORDS = {
+    "terminal", "run command", "open terminal", "execute command",
+    "dev terminal", "developer terminal", "run a command",
+}
+
+DANGER_TOKENS = {"rm", "rmdir", "sudo", "shutdown", "reboot", "dd", "mkfs", "kill", "pkill"}
+
+EXIT_WORDS = {"done", "bye", "goodbye"}
+EXIT_PHRASES = ["that's all", "close terminal", "never mind", "all done", "exit terminal"]
+
+CMD_SYSTEM_PROMPT = """You are a macOS terminal command generator. Convert the user's spoken request into a single shell command for zsh/bash.
+
+Rules:
+- Return ONLY the shell command — no explanation, no markdown, no backticks, no quotes
+- Compatible with macOS (zsh/bash)
+- Do not use sudo unless explicitly requested
+- Use conversation history to resolve relative references like "do it again" or "with verbose"
+
+Examples:
+"list files" -> ls -la
+"git status" -> git status
+"check disk space" -> df -h
+"find python files" -> find . -name "*.py"
+"recent commits" -> git log --oneline -10
+"what's running on port 3000" -> lsof -i :3000
+
+Return ONLY the command, nothing else."""
+
+SUMMARY_SYSTEM_PROMPT = """You translate raw terminal output into a short spoken reply.
+
+Rules:
+- 1 to 3 sentences, natural speech only
+- Extract the key facts; never paste raw output verbatim
+- If output is empty or blank, say the command finished with no output
+- If there is an error, say what went wrong in plain terms
+- Answer what the user actually asked for"""
+
+
+class VoiceDevTerminal(MatchingCapability):
+    worker: AgentWorker = None
+    capability_worker: CapabilityWorker = None
+
+    # Do not change following tag of register capability
+    # {{register capability}}
+
+    def does_match(self, text: str) -> bool:
+        t = text.lower()
+        return any(w in t for w in HOTWORDS)
+
+    def call(self, worker: AgentWorker):
+        self.worker = worker
+        self.capability_worker = CapabilityWorker(self.worker)
+        self.worker.session_tasks.create(self._run())
+
+    async def _run(self):
+        try:
+            trigger = await self.capability_worker.wait_for_complete_transcription()
+            history = []
+
+            inline = self._strip_trigger(trigger)
+            if inline:
+                await self._handle_command(inline, history)
+            else:
+                await self.capability_worker.speak("Terminal ready. What do you need?")
+
+            while True:
+                utterance = await self.capability_worker.user_response()
+                if not utterance:
+                    continue
+                if self._is_exit(utterance):
+                    await self.capability_worker.speak("Got it.")
+                    break
+                await self._handle_command(utterance, history)
+        except Exception as e:
+            self.worker.editor_logging_handler.error(f"[VoiceDevTerminal] {e}")
+        finally:
+            self.capability_worker.resume_normal_flow()
+
+    async def _handle_command(self, utterance: str, history: list):
+        cmd = self._generate_command(utterance, history)
+        if not cmd:
+            await self.capability_worker.speak("I couldn't generate a command for that.")
+            return
+
+        self.worker.editor_logging_handler.info(f"[VoiceDevTerminal] cmd: {cmd}")
+
+        if self._is_dangerous(cmd):
+            confirmed = await self.capability_worker.run_confirmation_loop(
+                f"That runs: {cmd}. Confirm?"
+            )
+            if not confirmed:
+                await self.capability_worker.speak("Cancelled.")
+                return
+
+        await self.capability_worker.speak("Running that now.")
+
+        try:
+            response = await self.capability_worker.exec_local_command(cmd, timeout=30.0)
+        except Exception as e:
+            self.worker.editor_logging_handler.error(f"[VoiceDevTerminal] exec error: {e}")
+            await self.capability_worker.speak("That command timed out or failed.")
+            return
+
+        output = response.get("data", "") if isinstance(response, dict) else str(response)
+        self.worker.editor_logging_handler.info(f"[VoiceDevTerminal] output: {output[:300]}")
+
+        history.append({"role": "user", "content": utterance})
+        history.append({"role": "assistant", "content": cmd})
+
+        spoken = self._summarize_output(utterance, cmd, output, history)
+        await self.capability_worker.speak(spoken)
+
+    def _generate_command(self, utterance: str, history: list) -> str:
+        raw = self.capability_worker.text_to_text_response(
+            utterance, history, CMD_SYSTEM_PROMPT
+        )
+        return (raw or "").strip()
+
+    def _summarize_output(self, utterance: str, cmd: str, output: str, history: list) -> str:
+        prompt = (
+            f"User asked: {utterance}\n"
+            f"Command run: {cmd}\n"
+            f"Output:\n{output or '(no output)'}"
+        )
+        raw = self.capability_worker.text_to_text_response(
+            prompt, history, SUMMARY_SYSTEM_PROMPT
+        )
+        return (raw or "Command completed.").strip()
+
+    def _is_dangerous(self, cmd: str) -> bool:
+        return bool(set(cmd.split()) & DANGER_TOKENS)
+
+    def _is_exit(self, text: str) -> bool:
+        t = text.lower().strip()
+        if any(p in t for p in EXIT_PHRASES):
+            return True
+        words = t.split()
+        return len(words) <= 2 and bool(set(words) & EXIT_WORDS)
+
+    def _strip_trigger(self, trigger: str) -> str:
+        cleaned = trigger.lower()
+        for hw in HOTWORDS:
+            cleaned = cleaned.replace(hw, "")
+        cleaned = cleaned.strip(" ,.-")
+        return cleaned if len(cleaned) > 3 else ""