Bot settings

Bot settings hold everything about one bot: who it is, what it reads, what it can reach, which model it runs on and how often it asks you. Open them with the Bot settings button in a bot's chat header. Settings shared by all your bots are in app Settings.

Use Search settings to filter the sections. Most changes save straight away; fields with a Save button wait for you. If you close with unsaved changes, Murage asks before discarding them.

Overview#

• What is this bot for?: choose Set up, say what you want the bot for, and pick a specialist profile or skills to match. If the bot is already set up, Murage tells you what it would change first.
• Appearance: the bot's Avatar.
  • Upload image: PNG, JPEG, GIF or WebP, up to 10 MB. You can also drop an image on the avatar.
  • Shape: Mascot, Circle, Rounded or Square.
  • Expression, Color and Body style the mascot. There are ten bodies.
  • Generate avatar makes one from an optional direction you type.
  • Reset mascot puts the default mascot back.
• Role: shown once you have more than one bot. Choose Chief of Staff (runs the whole workspace; one per workspace), Team leader (leads its team; one per team), Team member (works inside a team) or On its own (no team leader above it, reports to the Chief directly). An imported bot shows the role it came with, but importing never grants permissions.

See Bot identity and Face, voice and personality.

Identity & instructions#

• Name: up to 100 characters.
• Title: a short line on what the bot does, up to 200 characters.
• Instructions: what this bot does, what it should know and how you want it to work, up to 4,000 characters. The bot reads it at the start of every turn, and teammates read it when they decide who to hand work to, so keep it about the job.
• Personality: how the bot talks. Only the bot reads it; teammates deciding who to delegate to never see it.

What shapes this bot#

Shown with the bot's name, such as "What shapes Ember", in the desktop app. It lists everything that goes into the bot's instructions, in the order it reads them, grouped as Your rules, Who it is, What it can use and This turn.

Items you control have a switch or an Edit link: your House Rules, your About me profile, the Team brief, the bot's description and personality, its notes, and skills you gave it. Items marked Always on are Murage's own rules. Show exactly what it read shows the full text from the bot's last reply.

Skills

The skills this bot can open when a task needs one. Add a skill opens the library right here. Each skill can be read, switched on or off, or removed from this bot. A skill that needs a look first shows what the Skill Guard scan found before you can use it. See Skills.

Memory

• Managed memory: search what this bot remembers, see where each memory came from, and manage it. See Memory.
• Legacy notebook: editable Markdown notes from older versions. Import them into managed memory to make them available for recall.

Routines

The routines assigned to this bot, each marked Enabled or Paused. Open Routines takes you to the full routines view. Each routine can have its own approval level. See Routines and webhooks.

Access

• Connected apps: lets this bot use your connected Gmail, Calendar, Slack and other apps. Connect apps first under Tools → Connected apps. Some engines can't use them.
• App access and approvals: Only allow the accounts and actions I choose, then pick a Connected account and whether to Allow selected actions to send or change data. Choose Save app access. These limits cover Murage's connected apps only, not separate app logins, terminal tools or local files.
• Browser: gives this bot its own built-in browser tab, with its own logins, that you can watch and take over. It needs Built-in browser switched on under Settings → Experimental first. See Built-in browser.
• Computer: where this bot's computer runs: Cloud, Local VM, This computer or Off. With no choice made, it's on auto. For Cloud, pick a Cloud backend: Box (the default hosted computer) or Self-hosted VPS (your own Linux server over SSH). Start VPS automatically lets auto mode create or wake the bot's container when needed. See Computers.
• Working folder: where the bot runs its shell and file tools. Choose… a folder, or Clear to go back to the bot's private workspace. New tasks start in the new folder; a task already under way keeps its own.

Model

• Model: the default model for new threads. Existing threads keep their own.
• Effort: shown when the engine supports effort levels. Default lets the engine decide; the other levels depend on the engine.

See Engines.

Permissions

• Ask me before contacting other bots: when off, this bot contacts other bots without asking you first.
• Approvals: the default level for new conversations: Ask, Auto, Full access or No limits. New bots start on Ask. Full access and No limits can only be chosen in the desktop app, and the first time you choose one you confirm a warning. On Full access, a bot stops before messaging anyone new, posting publicly, paying, or deleting outside its folder, unless you have allowed it for that task. On Hermes and other ACP engines, Ask and the stop line cover only what the engine asks Murage about: shell commands and file edits. No limits turns that stop line off.
• Two extra switches apply only when the default is Full access or No limits:
  • Also skip approvals for my messages from Telegram, Slack and Discord. Messages from anyone else and webhooks still ask.
  • Also approve setup requests: installing skills, proposing routines and trusting folders. Connecting an app still asks.
• Always allowed: everything this bot has been told to always allow, including exact commands with their folder and engine. Remove takes any of it back. Desktop only.
• Routines use this level unless a routine has its own. Set a routine's level in its editor under Advanced → Approvals for this routine.
• Review routine approvals: Off (every approval waits for you), Watch (the same engine records a review without answering the card) or On (it answers only reviews that return a strict approval). Guardrail stops, unattended turns, access to this computer, and questions still wait for you. Not every engine supports this.

Full detail: Permissions and the stop line and the permissions matrix.

Voice & alerts

• Voice: this bot's own voice for calls and spoken replies. Pick a Voice engine: Flux (or OpenAI, when you use your own OpenAI key), xAI, ElevenLabs (with your ElevenLabs key), or the built-in Mac or Windows voices. Then pick a voice. Every voice in the list has its own play button. See Voice providers.
• Read replies aloud: speak this bot's answers as they arrive, even from another chat.
• Notifications: get notified when this bot finishes or needs input.

History

This bot's tasks, with the date each started. Choose one to open it.

Usage

Turns, Tokens and Cost for this bot. On a subscription engine the cost is shown as an equivalent; engines that report no price show tokens only. All bots opens the Usage tab in Settings. See Usage and cost.