The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Choose a Hermes Agent feature by asking what must persist, what should stop the work, and whether you need more workers or more model perspectives. Use Loop for repeated turns in the active session, Goal for continuing toward one judged outcome, Cron for scheduled work outside an interactive session, delegation for temporary parallel subtasks, Kanban for durable task tracking and handoffs, Bot Mode for reusable specialist profiles, and Mixture of Agents for model perspectives synthesized by one acting agent.
How do the Hermes Agent features differ?
They address different parts of an agent workflow rather than offering seven interchangeable ways to ask for help. Some control when work runs, some organize who does it, and one changes how models contribute to a response.
As an Amazon Associate I earn from qualifying purchases.
| Feature | Best fit | Where work persists | How it ends or progresses |
|---|---|---|---|
| Loop | Repeat a check or prompt during an active session | Current session | Runs again on a cadence while the loop is active |
| Goal | Keep pursuing one defined outcome | Continuation across turns in the current interaction | A judge checks completion; continuation is subject to a turn budget and controls |
| Cron | Run work unattended on a schedule | Scheduled job outside an interactive session | Runs at its scheduled time or interval; jobs can also be event-triggered |
| Delegation | Split a bounded task into independent parallel subtasks | Temporary child-agent runs; summaries return to the parent | Children finish their subtasks and report back |
| Kanban | Manage multi-stage work, handoffs, or multiple profiles | Durable shared board and task records | Workers and handoffs advance the tracked work |
| Bot Mode | Reuse named specialists with distinct configurations | Named profile configuration | Used when the relevant Bot is invoked or assigned work |
| Mixture of Agents | Bring several model perspectives to one response | One routed turn, not a persistent task board | Reference models provide analysis; an aggregator produces the response and tool calls |
The key distinction is whether you need a cadence, an outcome, a schedule, temporary parallel workers, durable coordination, reusable profiles, or model-level synthesis.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallWhen should you use Loop?
/loop repeats a prompt or another slash command on a recurring cadence inside the active session. Each wakeup is a real agent turn that reads the current state again, so it can suit polling a deployment, checking a queue, or repeating housekeeping while you are working.
Choose a fixed interval when an external schedule matters; choose a self-paced loop when progress should determine the rhythm. A loop is session-bound. For work that needs to run unattended or survive terminal restarts, use Cron instead.
When is Goal better than a loop?
/goal <text> sets a standing objective. After each turn, an auxiliary judge model checks whether the goal appears complete. If it is not complete, Hermes can continue, subject to the configured controls and turn budget. The slash-command reference documents a default budget of 20 turns; defaults can vary by software version, so check the installed version before relying on that number.
Use Goal when the stopping condition is an outcome, such as finishing a bounded implementation task. A loop is about repeating at a cadence; a goal is about continuing until completion is judged. User messages preempt Goal continuation, so it is not the same as an unattended scheduled job.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesWhen does Cron make more sense?
Cron is the option for work that should run at a specified time or interval outside an interactive session. It supports one-shot and recurring schedules, as well as event-triggered jobs. Jobs can be paused, resumed, edited, run manually, or removed.
A Cron job can start a fresh agent session or run a no-agent script whose output is delivered verbatim. It can use attached skills and deliver results to the originating chat, files, or configured platform targets. Model selection can be pinned per job, inherited from a Cron-wide default, or resolved from the main model when the job fires.
This makes Cron a better fit than Loop for recurring work that must continue when the interactive session is no longer active. It is also distinct from Goal: Cron decides when work runs, while Goal is a way to continue toward an objective.
What does delegation add?
delegate_task launches child agents with isolated context and terminal sessions. The parent can continue working while children handle independent, bounded subtasks; their final summaries return to the parent for synthesis. The documented default is up to 10 concurrent subagents in a parallel batch, configurable by the user. That is a product default, not a performance guarantee.
Free tools Windows power users keep installed
One-click scans. No signup required.
Delegation is useful when splitting work saves time or allows separate lines of inquiry to proceed independently. It is less suitable for work that needs durable tracking or several explicit handoffs. Child terminal processes belong to the child, so manage any process that must persist in the parent or through a durable mechanism.
How is Kanban different from delegation?
Kanban provides a durable task board shared across profiles, with task and handoff records and workers that run as separate processes. It is aimed at workflows that need their state to remain visible and available beyond a temporary parent-child run, such as multi-stage pipelines, research triage with a human in the loop, recurring operations journals, or persistent named assistants.
Delegation is a temporary way to launch child workers and bring their summaries back to a parent. Kanban is for retaining task state and coordinating handoffs. A Kanban board is not simply another name for a delegation call.
What does Bot Mode configure?
Bot Mode lets you create named Bots with configurable model or provider, profile instructions, skills, toolsets, MCP servers, and separate routines. Its purpose is to make specialists reusable and easier to access.
A Bot answers “which configured specialist should do this?” It does not, by itself, provide a durable multi-stage task workflow. A broader setup can use named Bots alongside Kanban for persistent coordination or delegation for temporary child workers; those features address different needs.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What happens in Mixture of Agents?
Mixture of Agents (MoA) is a virtual model provider. Reference models contribute analysis, then an aggregator model writes the response and emits tool calls. The aggregator runs the tool loop and accounts for most of the run’s cost. The /moa shortcut routes one prompt through the default preset and restores the previous session model afterward.
MoA changes how a turn gathers and combines model perspectives; it does not create task-owning child agents or a durable board. If evaluating cost, the aggregator’s provider is the central consideration described in the documentation.
Which feature should you start with?
| If you need to… | Start with | Because… |
|---|---|---|
| Repeat a check while the conversation remains active | Loop | It creates recurring turns within the current session. |
| Keep working toward one bounded result | Goal | Continuation is based on a judged outcome, subject to controls and a turn budget. |
| Run a job unattended at a time or interval | Cron | It schedules work outside an interactive session. |
| Split independent subtasks and combine their results | Delegation | Temporary children work in isolated contexts and return summaries. |
| Track stages and handoffs across profiles | Kanban | It retains task state on a shared durable board. |
| Reuse specialists with different settings | Bot Mode | It configures named profiles with their own instructions and capabilities. |
| Get several model perspectives for one response | Mixture of Agents | Reference models advise an aggregator that remains the acting agent. |
Can these features be combined?
Yes, when each feature has a distinct role. For example, a multi-stage project might use named Bots for recurring specialist configurations, Kanban to retain tasks and handoffs, and delegation for a bounded parallel investigation. If a check needs to run on a schedule outside the session, Cron is the relevant scheduling layer. MoA may be used when a single response benefits from reference-model analysis and an acting aggregator.
Before choosing, define the work’s stopping condition and persistence requirement. If it should stop when an outcome is judged complete, start with Goal. If it should run on a schedule, start with Cron. If it needs durable ownership and handoffs, start with Kanban. If independent subtasks can finish and report back in one run, delegation is the lighter coordination shape.
Version and behavior checks
Hermes Agent features, command behavior, and defaults are software-version-sensitive. The documentation reviewed on October 7, 2026 describes the behaviors above, but verify the installed version’s command reference and configuration before relying on a default such as Goal’s turn budget or delegation’s concurrency limit.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
