Skip to content

Agent Hub

AI you deploy — and govern

The Agent Hub is your command center for AI that reads your data, surfaces insight, and takes action — inside guardrails you set, on the model you choose.

Browse the Marketplace Build a custom agent

What an agent is

An agent is a named AI teammate with a job. It works over the same single record the rest of Muntri uses — accounts, deals, activity, usage, conversations — so it always reasons on live, joined-up data. Every agent has an identity (a name and a friendly avatar), a set of capabilities, and a governance policy that decides what it may do on its own versus what needs your sign-off.

1shared record
3autonomy tiers per action
∞agents you can run
Anymodel you bring

The kinds of agent

Muntri has a small family of agent types that work together — from one-click templates to a supervised org that spans your whole go-to-market.

  • Marketplace agents

    Pre-built templates across sales, CS, ops and marketing. Activate one and it becomes your own live, editable agent.

  • Custom agents

    Describe what you want in plain language, then wire the data sources, schedule and outputs. Yours end to end.

  • SDR agents

    Research prospects and draft outbound email, so reps start every conversation informed.

  • Master Agents → Process Agents

    A Master Agent supervises a set of Process Agents — the visual workflows that run your repeatable plays.

  • Customer agents

    A persistent per-account agent that carries the context of one relationship and lets owners tune the automation for that account.

  • User agents

    Your personal copilot. Rename it, re-skin it, and use "Ask {agent}" to query your data anywhere.

The agent org

Master Agents sit at the top as supervisors. Each one manages a set of Process Agents — versioned visual workflows that run your plays (qualify a lead, progress a deal, triage a ticket, watch a health score). Alongside them sit the per-account Customer agents and per-user User agents that hold ongoing context.

The Agents home brings every agent into one place. Switch between Lifecycle (your agents laid out left-to-right along the revenue journey, from demand through prospecting, deals, onboarding, retention and expansion), Hierarchy (the master-to-process org chart) and List views. Click any agent to see what it does, its live impact and its recent activity — and jump straight to managing it.

flowchart TD
  M[Master Agent<br/>supervisor] --> P1[Process Agent<br/>Leads]
  M --> P2[Process Agent<br/>Deals]
  M --> P3[Process Agent<br/>Support]
  CA[Customer agents<br/>per account]
  UA[User agents<br/>per person]
  M -.oversees.- CA
  M -.oversees.- UA
  classDef n fill:#ffffff,stroke:#2563eb,stroke-width:1.5px,color:#0B1024;
  class M,P1,P2,P3,CA,UA n;

Identity: names & avatars

Every named agent shows a friendly portrait rather than a bare initial, so your fleet feels like a team. Pick a face from the built-in avatar library on any agent, and rename your personal assistant whenever you like — from the web app or on mobile.

Make it yours

Open your assistant's identity, set a name and choose an avatar. The new name and face follow you across every screen, including the "Ask {agent}" chat.

Ask

Every user has a personal copilot in the app chrome — the Ask {agent} pill. Ask a question in plain language and it answers over your live tenant data, with voice in and read-aloud on mobile. Your conversation is kept and synced across devices, so you can pick up where you left off.

  • Ask in plain language

    "Which of my accounts are at risk this quarter?" — answered from your real data.

  • Voice on mobile

    Speak your question and have the answer read back to you, hands-free.

  • Continuous thread

    One conversation, synced across your devices.

What your agent knows — and its impact

Open any Customer or User agent to see the living brief it maintains about that account or person. Alongside it sits an Impact panel — a live read-out of what the agent is tracking and influencing (deals and revenue in play, health and sentiment signals, risks flagged, and the agent's own work) — plus a history showing how its understanding has evolved, version by version.

Grounded in your data

Every number on the Impact panel is derived from your live tenant data, so it always reflects the current state of the account or rep.

The governance model

Agents are useful because they act. They're safe because you decide how they act. Two controls work together.

Guardrails

Set the boundaries an agent operates within — what it may touch and what it must never do. Guardrails travel with the agent so its behavior stays predictable as it runs.

Per-action autonomy

Beyond broad guardrails, you set a tier for each kind of action a Process Agent can take. There are three:

Tier What happens Use it when
Auto The action runs on its own. You trust the play and want speed.
Require approval The run pauses and notifies you; it proceeds only after you approve. The action is high-stakes or customer-facing.
Advisory Nothing is executed — a suggestion is logged for you to act on. You're still building trust in the play.

Start conservative, watch the results in AI Activity, then promote actions to a higher tier as your confidence grows. Full detail: Running agents & autonomy.

Bring your own model

Muntri is model-agnostic. Point your agents at the provider you prefer and run inference at that vendor's token cost.

  • Claude · GPT · Gemini · Llama

    Choose a leading provider and the specific model per feature, with optional automatic failover to a backup.

  • Your own endpoint

    Route to a self-hosted or private model when you need full control.

How model choice works

Set a default provider for your workspace and, optionally, route specific AI features to specific models. Usage is metered so you can see spend by model, and you can cap it. When you bring your own model, inference runs at your vendor's token cost.

Where to go next