task.list
One sidebar, all your tasks
Long-running servers, watchers, build steps — defined in plain JSON. Started and stopped from a tree view in VS Code, Cursor, or Windsurf.
One place for your dev tasks.
One place for your logs.
And your AI agent sees them too.
A small VS Code extension that gives MCP-enabled agents — Claude Code, Cursor, Windsurf — a real handle on your long-running dev processes. Start, stop, restart, read logs. No hidden terminals. No retry loops.
VS Code CLI
code --install-extension Tolbxela.taskdev
Or in-editor
Extensions panel → search "TaskDev"
Then drop a taskdev.json in your workspace and run
TaskDev: Install MCP config from the command palette to wire up your agent.
// what it does
task.logs
Logs your agent can actually read
Every task writes to a real file. Your editor opens it; the agent reads the same file over MCP. No hidden terminals, no scraped output.
mcp.tools
Eight MCP tools, sandboxed
list, status, control, restart, logs, history, add, remove. Agent-added commands are sandboxed: no shell chaining, no traversal, only known dev shapes.
privacy.local
Local. No telemetry.
Tasks run as local processes. Logs stay on disk. Nothing is shipped anywhere. MIT-licensed, source on GitHub.
// read the write-up
The story behind TaskDev — the agent problem and how MCP fixes it — on Medium and dev.to.