원클릭으로
dev-server
Start Tandem dev environment (server + client) and verify MCP connection
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
메뉴
Start Tandem dev environment (server + client) and verify MCP connection
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Cut a Tandem release — six-surface version bump, changelog, tag, GitHub Release publish, smoke checklist
Use when tandem_* MCP tools are available, the user asks about Tandem document editing, or iterating on text collaboratively. Provides workflow guidance, annotation strategy, and tool usage patterns for the Tandem collaborative editor.
Generate a Keep a Changelog entry from git log since the last tag
Capture README screenshots of the Tandem editor UI via Playwright + MCP
Debug Playwright E2E test failures — port conflicts, server startup, test isolation, and post-mortem analysis
Run Playwright E2E tests safely (warns about dev server conflicts)
| name | dev-server |
| description | Start Tandem dev environment (server + client) and verify MCP connection |
| disable-model-invocation | true |
Start the Tandem development servers and verify everything is connected.
freePort() kills any existing process on :3478/:3479 at startup — you cannot run two instancesfreePort() does NOT cover :5173 — an orphaned Vite from a killed dev:standalone survives, and because vite.config.ts sets strictPort: true the new Vite exits with an error rather than falling back to another port. The symptom is a dev server that won't start, not a silently-wrong URL (see step 2)npm run test:e2e) will also kill dev servers on those portsRun the doctor first — it diagnoses Node version, .mcp.json, server health, and ports before anything else:
npm run doctor
If you pulled since the last dev-server run, also refresh dependencies (they drift):
npm install
Check if servers are already running:
curl -s http://127.0.0.1:3479/health 2>/dev/null && echo "Server already running" || echo "Server not running"
If the server is NOT running but something still holds :5173, it's an orphaned Vite from a previous dev:standalone. Kill it first — strictPort: true means the new Vite will otherwise fail to start with Port 5173 is already in use instead of picking another port.
Windows (git-bash) — note the PID in the last column:
netstat -ano | grep ":5173" | grep LISTENING
taskkill //PID <pid> //F
macOS/Linux:
lsof -ti tcp:5173 | xargs -r kill
If not running, start the standalone dev environment (server + client):
npm run dev:standalone
This starts:
:3478:3479:5173Wait for both servers to be ready:
# Wait for MCP health endpoint
for i in {1..15}; do curl -sf http://127.0.0.1:3479/health && break || sleep 1; done
Verify MCP connection works by running /mcp in Claude Code if needed.
Open the editor at http://127.0.0.1:5173 in the browser.
If you changed server code, restart just the server:
dev:standalone processnpm run dev:standalone again/mcp in Claude Code to reconnect