Platform Features
This page describes what Mutiro can do, how each feature works from the user's perspective, and which clients support it.
For details on each client (download links, how to connect), see clients.
Key distinction: Mobile and terminal chat are messaging clients only. The web app supports messaging plus creating and managing Mutiro-hosted agents. Desktop and CLI add full self-hosted agent management on top — running local daemons, editing runtime config, swapping the brain.
Messaging
All clients support these messaging features:
- Text messages — send and receive text
- Voice messages — record audio (mobile/desktop), send audio files (CLI). The agent receives a text transcription automatically
- Image messages — send and view images
- File messages — send and receive any file type
- Reactions — tap/click to react with emoji
- Forwarding — forward messages to other conversations
- Reply — reply to a specific message in a thread
- Read receipts — messages are marked as read automatically
- Live activity signals — you see what the agent is doing in real time (thinking, recalling, searching, reading files, writing, etc.)
Message search
- Desktop: yes (search within conversations)
- CLI: yes (
mutiro user message search <query>) - Mobile: not yet
- Web: same as desktop
Voice Messages
Sending voice (you to agent)
Record audio in the mobile or desktop app. The audio is transcribed to text automatically — the agent receives the text, not the audio file. You'll see a "transcribing" indicator while it processes.
From CLI: mutiro user message send-voice <audio-file>
Receiving voice (agent to you)
When an agent sends a voice message, you hear synthesized speech. The agent chooses when to respond with voice vs text. You can play it back in any client.
The agent's voice and language are configured by the agent owner (not by you).
Conversations
- New conversation — start a chat with any user or agent by searching for their username
- Conversation list — see all your conversations, sorted by recent activity
- Presence — see who's online (green dot = online)
Available on all clients.
Interactive Cards
Agents can send interactive elements in the chat — forms, polls, checklists, buttons, and more. You interact with them by clicking, filling in fields, and submitting. The agent sees your responses and can update the card or reply.
Available on: Desktop and web app. Mobile shows card content but with limited interactivity. CLI does not render cards.
You don't need to do anything special — cards appear as normal messages in the chat. The agent decides when to send one.
Scheduling
Agents can schedule tasks and reminders. You interact with scheduling through normal conversation:
- "Remind me tomorrow at 8am to call Mom"
- "Check the deploy status every hour"
- "Send me a weekly summary every Monday morning"
The agent handles it internally. When the scheduled time arrives, the agent's response appears in your chat like a normal message. There is no separate "schedules" screen or UI — it's all conversational.
To use scheduling: The agent must have the scheduling tools enabled (schedule_message_create, schedule_message_list, schedule_message_cancel). These are enabled by default.
Timezone: The agent respects your local timezone. If you say "8am," it means 8am in your timezone.
Memory
Agents remember things across conversations. There are three layers, but from your perspective it's seamless — the agent just remembers.
What happens behind the scenes
- Long-term memory — The agent saves important facts (your preferences, decisions, context) and recalls them in future conversations. This persists forever until the agent owner clears it.
- Working memory — The agent tracks the current conversation's goals, open questions, and context. This resets when the conversation changes topic.
- Recall — The agent can search through earlier messages in the conversation to find things you discussed previously, even if they've scrolled out of the current context.
What you experience
The agent remembers your name, preferences, past decisions, and ongoing projects without you repeating yourself. There is no UI to browse or edit what the agent remembers — it's automatic and managed by the agent.
To use memory: The agent must have memory tools enabled (memory_get, memory_write). These are enabled by default.
Live Calling
Voice calls with agents in real time. The agent speaks and listens during the call, not through text messages.
Available on: Mobile and desktop. Not available on CLI or TUI.
To use live calling: The agent owner must enable live calling. For hosted agents this is managed through platform settings; for self-hosted agents it is live_calling_enabled in the agent config.
Agent Management
Mutiro-hosted agent creation and management is available on Desktop, Web, and CLI. Self-hosted agent management — running local daemons, editing runtime config, owner-only tool toggles — is Desktop and CLI only.
Desktop App
The desktop app has a full visual interface for managing agents:
- Create agent — fill in username, display name, and objective
- Agent list — see all your agents with online/offline status
- Runtime status — view hosted status and start/stop self-hosted daemons
- Edit profile — change display name, bio, avatar
- Configure hosted agents — manage owner-facing settings such as instructions, voice/language, profile, and sharing; hosted tool settings use catalog defaults
- Configure self-hosted agents — visual editor for owner-managed runtime wiring:
- LLM provider and model selection
- Tool enablement with owner-only toggles
- Voice and language settings
- Live calling toggle
- Workspace configuration
- Model parameters (temperature, max tokens, etc.)
- Manage allowlist — add/remove users who can message the agent
- API key — regenerate API keys
- Runtime — manage multiple agents simultaneously
Web App
The web app supports the hosted agent lifecycle from a browser:
- Create hosted agent — fill in username, display name, and objective
- Agent list — see your hosted and self-hosted agents
- Edit hosted profile — display name, bio, avatar
- Configure hosted agents — instructions, voice/language, sharing
- Manage allowlist — add/remove users who can message the agent
Self-hosted agents created from desktop or CLI show up in the list, but their runtime configuration (LLM provider, tool toggles, daemon start/stop) stays on the desktop app and CLI — the web app has no daemon to drive.
CLI
The CLI provides agent management commands. Hosted agents use platform-managed runtime settings; self-hosted agents can also use runtime config files and daemon commands in the environment where you run them.
What you CANNOT do on mobile
- Create, delete, or configure agents
- Start or stop agents
- Edit agent profiles or config
- Manage allowlists
- Manage API keys
Mobile is purely for messaging. Use the web app, desktop app, or CLI for agent management.
What you CANNOT do on web
- Create or run self-hosted agents (no daemon in a browser)
- Edit self-hosted runtime config — LLM provider, model parameters, tool toggles
- Start or stop self-hosted daemons
- Manage API keys for self-hosted agents
Hosted-side actions — create, configure, share, allowlist — work on the web app. Anything that drives a local daemon needs the desktop app or CLI.
Feature Matrix
| Feature | Mobile | Desktop | Web | CLI | Terminal Chat |
|---|---|---|---|---|---|
| Send/receive text | Yes | Yes | Yes | Yes | Yes |
| Voice messages | Yes | Yes | Yes | Yes (file) | No |
| Image messages | Yes | Yes | Yes | Yes | No |
| File messages | Yes | Yes | Yes | Yes | No |
| Reactions | Yes | Yes | Yes | Yes | No |
| Forwarding | Yes | Yes | Yes | Yes | No |
| Message search | No | Yes | Yes | Yes | No |
| Interactive cards | Basic | Full | Full | No | No |
| Live calling | Yes | Yes | No | No | No |
| Live activity signals | Yes | Yes | Yes | No | No |
| Presence | Yes | Yes | Yes | Yes | Yes |
| Create agents | No | Yes | Hosted only | Yes | No |
| Configure agents | No | Yes | Hosted only | Yes | No |
| Start/stop self-hosted daemons | No | Yes | No | Yes | No |
| Manage allowlist | No | Yes | Yes | Yes | No |
| Scheduling | Conversational | Conversational | Conversational | Conversational | Conversational |
| Memory | Automatic | Automatic | Automatic | Automatic | Automatic |
For self-hosted tool configuration and security implications, see configuration and sharing and security. Mutiro-hosted agents use hosted catalog defaults.
What Doesn't Exist Yet
To be clear about current limitations:
- No agent creation on mobile — use web app, desktop app, or CLI
- No self-hosted agent management on web — hosted agents work in the browser; self-hosted agents need desktop app or CLI
- No schedule management UI — scheduling is conversational only (talk to the agent)
- No memory browser — you can't see or edit what the agent remembers
- No agent-to-agent messaging — agents can't message each other directly
- No group conversations with multiple agents — one agent per conversation
- No message search on mobile — available on desktop and CLI only