SILO Task Control
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@SILO Task ControlOpen the control panel and scan my current tasks."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
SILO Task Control
SILO is a lightweight Codex control panel for scanning existing tasks across projects, reviewing the exact Continue or Optimize prompt for each task, and dispatching a confirmed batch through Codex's native task tools.
Current release: v0.3.1 (0.3.1+codex.20260913)
Install
Add the GitHub repository as a Codex marketplace, then install SILO:
codex plugin marketplace add LIMITEDRUSH/silo-task-control
codex plugin add silo-task-control@silo-pluginsRestart Codex after the first installation. Open SILO from the sidebar or start a new task and ask:
$silo-task-control Open the control panel and scan my current tasks.SILO asks you to choose 简体中文 or English the first time its panel opens. The choice controls both panel copy and generated task prompts, is remembered locally, and can be changed from the language button in the header.
Update later with:
codex plugin marketplace upgrade silo-plugins
codex plugin add silo-task-control@silo-pluginsSILO currently targets Windows and requires Codex Desktop plus Node.js 22.5 or newer.
Related MCP server: LightWorker
Highlights
Fast local task inventory with Codex-native running-state synchronization and a SQLite fallback.
Minimal All / To do / Completed / Smart select workflow with global search and project grouping.
Exact per-task prompt review, Continue/Optimize mode, independent Burn strategy, model and reasoning controls, and explicit launch confirmation.
A live activity rail backed by durable native jobs, atomic target claims, and progress restoration after the panel or server restarts.
Lightweight quota polling and reset prediction. Reset information is fetched from the third-party codex-reset.com service when the panel opens.
Keyboard navigation:
/focuses search, arrows move between visible tasks,Xtoggles selection,Spaceopens details, andEscclears the batch without deleting drafts.
How dispatch works
Opening, scanning, filtering, and Smart select are read-only. Smart select reviews at most the latest two turns of recent candidates and proposes Continue or Optimize without launching anything.
Launch requires an explicit review dialog. SILO persists a short-lived plan, sends a visible follow-up message, re-reads each target through native Codex task tools, skips active or attention-blocked tasks, atomically claims each eligible target, and records progress in the activity rail. It never creates, forks, archives, renames, or interrupts unrelated tasks.
External Windows panel
Choose External window in SILO, or double-click desktop\Start SILO.cmd. The launcher uses the Node and Codex runtimes bundled with Codex Desktop, starts a loopback-only service, and opens the same panel in an Edge application window.
Sidebar entry
SILO advertises its MCP App as a global entry point, allowing compatible Codex Desktop builds to show SILO in the sidebar. This metadata is still an undocumented host behavior. After a Codex Desktop upgrade, run:
npm run sidebar-doctorThe doctor performs a read-only compatibility check. SILO does not patch Codex or modify app.asar.
Development
npm install
npm run checkThe bundled MCP server is dist/server.mjs.
Safety
Scans and plan creation are read-only.
Plans expire after 15 minutes and contain only previously scanned task IDs.
Dispatch requires explicit confirmation and produces a visible Codex message.
Edited prompts are sent exactly as reviewed; Burn is appended once when enabled.
Every target is checked again immediately before dispatch.
The desktop STOP control affects only child tasks launched by that SILO desktop process.
License
MIT
This server cannot be deployed
Maintenance
Related MCP Connectors
AI work orchestration for plans, tasks, teams, and coding-agent dispatch.
- DazbenchOAuthapp.dazbench
Task management your AI agents can actually run. One line becomes a context-ready task over MCP.
Connect, monitor, and control AI agents — tasks, approvals, schedules, and governance.
- projectsOAuthcloud.tri2b
Task tracking built for coding agents. Work is leased, so two agents never take the same SubTask.
Related MCP Servers
- AlicenseCqualityDmaintenanceEnables Codex to delegate decision-making to an OpenAI reviewer for structured next-step recommendations during long-running tasks.1MIT
- AlicenseAqualityAmaintenanceLocal-first multi-agent delegation and approval control for Codex via MCP, with persistent task DAG, isolated worktrees, and a web console.141MIT
- AlicenseNot gradedqualityBmaintenanceEnables running configurable Prompt-as-Code tasks across multiple scenarios through local Codex, with MCP tools to list scenarios, execute standard task files, and validate task definitions.MIT
- FlicenseBqualityBmaintenanceEnables local, stdio-based task dispatch from Codex to DeepSeek Harness with independent review, bounded rework, persistent queues, and completion tracking.15-