Hermes Agent Mission Control

What is Hermes Agent Mission Control?

Hermes Agent Mission Control (Minions) is a “Mission Control” Kanban interface for the Hermes Agent. It’s designed to help users supervise autonomous Hermes Agent work by turning multiple long-running agent sessions into a single, durable board with task states and review checkpoints.

Instead of manually juggling terminal sessions and checking progress one by one, Minions tracks each task as a persistent Hermes session, logs heartbeat progress, streams tool activity, and keeps task metadata and status in a local SQLite database.

Key Features

• Kanban board with task status: View tasks at a glance across phases such as in progress, blocked, in review, and done.
• Autonomous execution workflow: Describe what you want in chat; the agent decides how to complete the task.
• Heartbeat check-ins: Agents periodically report progress. Blocked work surfaces automatically, and the session can retry with a different approach before asking for help.
• Live streaming during execution: Watch tool calls, reasoning, and responses in real time as the agent runs.
• Human-in-the-loop completion control: Agents propose completion; you verify and close tasks, and nothing moves to done without your sign-off.
• Per-task model and reasoning effort control: Override model and reasoning effort per task via task-specific settings.
• Cron job visibility: Track scheduled Hermes jobs, including history and output.
• Local-first self-hosted option: Run locally with SQLite—no account and no cloud dependency mentioned—so your local data stays on your machine.

How to Use Hermes Agent Mission Control

1. Install prerequisites: Use Node.js 18+ and have Hermes Agent available as expected by the repository.
2. Run the app locally:
   • git clone https://github.com/Agent-3-7/hermes-agent-mission-control.git
   • cd hermes-agent-mission-control
   • npm install
   • For development: npm run dev
3. Open the UI: Navigate to http://localhost:6969.
4. Run production (if desired): Use npm run prod.

Internally, the app uses a React + Vite browser UI connected to an Express server on port 6969 using HTTP and SSE for streaming updates. A Python worker interfaces with Hermes via the Hermes AI component, and task transcripts are stored in Hermes’s session database while Minions stores task metadata, status, heartbeat history, and per-task settings in local SQLite.

Use Cases

• Reviewing multiple delegated agent tasks: When you delegate research, coding, monitoring, sales ops, writing, or recurring workflows, Minions provides a single board to see which tasks are in progress, blocked, or ready for review.
• Handling stuck or blocked executions: If a task repeatedly stalls, the heartbeat mechanism surfaces it to you as “Needs your help,” after the agent has tried alternative approaches.
• Monitoring long-running work without constant manual checking: Live streaming plus periodic heartbeats reduce the need to repeatedly open and inspect terminal sessions while work is underway.
• Operating scheduled Cron-based Hermes jobs: Use the cron visibility features to see scheduled jobs, inspect history, and view output when tasks are triggered in the background.
• Running per-project tasks with different execution settings: Override model and reasoning effort on a task-by-task basis, which is useful when different jobs have different constraints.

FAQ

• Can I use this with other agents besides Hermes?
  Not yet. The source notes that an adapter interface exists, but the launch is Hermes-only and calls out OpenClaw as next.

• Where is task data stored?
  The project describes a local-first option using SQLite for task metadata, status, heartbeat history, and per-task settings. Chat transcripts are stored in Hermes’s session database.

• How does “done” work?
  The agent can propose completion, but you must verify and close tasks. The board only moves tasks to done after your sign-off.

• Does Minions only show a board or does it actively manage execution?
  It does more than display states: each task is a persistent Hermes root session that talks, checks in, and updates the board based on heartbeat events.

Alternatives

• General-purpose Kanban/task trackers: Tools that track statuses manually can help organize work, but they typically don’t provide agent-specific heartbeat check-ins, streaming of tool calls/reasoning, or per-task execution control.
• Terminal-based Hermes session management: Running Hermes directly in terminals can work for small numbers of tasks, but it reintroduces the manual effort Minions is meant to reduce (tracking completion and diagnosing blocked work).
• Self-hosted orchestration dashboards for background jobs: These can provide visibility into scheduled jobs and execution logs, but may not offer the Hermes-focused task/session model with human-in-the-loop “ready for review” workflows described here.