Skip to content

Agent Hub

Running agents & autonomy

Decide, action by action, what your agents may do on their own — then watch every outcome and approve the rest in one tap.

Agent Hub overview Build a custom agent

Autonomy, one action at a time

The most important control in the Agent Hub is per-action autonomy. Rather than a single on/off switch, you set a tier for each kind of action a Process Agent can take. This lets you move fast on the plays you trust while keeping a human in the loop on the ones that matter.

Tier What happens Best for
Auto The action runs automatically as part of the workflow. Routine, low-risk steps you've seen work.
Require approval The run pauses and notifies you; it continues only after you approve — and then re-runs that exact step. High-stakes or customer-facing actions.
Advisory Nothing runs — the agent logs a suggestion for you to act on manually. New plays you're still building trust in.

A safe way to roll out

Start new actions as advisory, review the suggestions in AI Activity, then move the good ones to require approval, and finally to auto once you're confident. You set these tiers once and they apply across your Process Agents.

How an action passes the gate

flowchart TD
  S[Agent reaches an action] --> G{Autonomy tier?}
  G -- Auto --> R[Run it]
  G -- Require approval --> P[Pause & notify]
  P --> Q{You decide}
  Q -- Approve --> R
  Q -- Reject --> X[Skip]
  G -- Advisory --> L[Log a suggestion]
  R --> H[Recorded in AI Activity]
  L --> H
  X --> H
  classDef n fill:#ffffff,stroke:#2563eb,stroke-width:1.5px,color:#0B1024;
  class S,G,R,P,Q,X,L,H n;

The approval inbox

When an action is set to require approval, the run pauses and you get a notification. Open the pending item, review exactly what the agent wants to do, and approve or reject it. On approval, the agent resumes and carries out that step; on rejection, it moves on without acting.

  • You're notified

    A paused, approval-required action pings you the moment it's ready for review.

  • One-tap on mobile

    Approve or reject from the Muntri mobile app — push straight to your phone, so agents never stall waiting on you.

  • It resumes on approval

    Approve and the agent completes the exact step it paused on, then continues the run.

Run history & outcomes — AI Activity

Every agent run, recommendation and action lands in AI Activity — a legible feed of what your AI actually did. Use it to see outcomes, spot suggestions waiting on you, and build the confidence to raise an action's autonomy.

  • What ran

    Workflow runs, agent actions, drafted emails and generated recommendations, in one timeline.

  • What's suggested

    Advisory suggestions and pending approvals surfaced so nothing is missed.

  • What it cost

    AI usage and spend so you can keep an eye on consumption by feature.

Running agents on your terms

  • On demand

    Expand any agent in the Fleet and run it now to see fresh output immediately.

  • On a schedule

    Give an agent a recurring cadence so it works without you having to trigger it.

  • Event-driven

    Process Agents run when something happens — a lead is created, a deal changes, a health score drops.

Scheduled reports & digests

Give an agent a schedule and it delivers on its own — a daily, weekly or hourly digest sent straight to email and Slack, with no one pressing run.

  • In your timezone

    Pick the time and timezone. Runs land at the right local hour all year — daylight saving is handled for you.

  • Email + Slack

    The same report goes to the recipients and the Slack channel you choose.

  • Readable at a glance

    Reports arrive in a clean, light layout led by a graphical KPI header — trend cards comparing the last 24 hours against the prior day, 7-day and 30-day with up/down deltas and mini charts — then the written analysis and the underlying data.

Choose a report style

Give each agent a report style — pick a preset (Executive Brief, Analytics Deep-Dive, Visual Dashboard, Newsletter, or Standard) and tune the tone, length and which sections appear. Set a workspace default so every agent's report follows your house style unless it overrides it.

Guardrails still apply

Autonomy tiers decide whether an action runs, pauses or logs. Guardrails decide the boundaries the agent works within in the first place. The two work together — set both when you build or activate an agent.