Mutiro Manual

Start here

What is Mutiro — how the platform works, key concepts, who it's for
Agents, Conversations, and Workspaces — how one agent serves many people; context, memory, shared files, and owner/developer access
Getting Started — sign up and create your first agent
Connecting to Your Agent — web, mobile, desktop, terminal, CLI
Features — what's available on each client, feature matrix
Scheduling — send a message later or schedule a prompt that asks the agent to do fresh work
Plans and Limits — free-tier limits and the paid plans that raise them
Sharing and Security — allowlist, workspace isolation, threat model
Troubleshooting — common issues and fixes

Step-by-step walkthroughs and worked examples — read these alongside the reference docs above.

Web quickstart — illustrated, no install: sign up and create your first hosted agent in the browser
CLI quickstart — the same path through mutiro commands
Self-hosted quickstart — run the agent daemon yourself
Build agents with your coding assistant — give Claude Code, Codex, or Cursor the Mutiro developer guide as a plugin or skill, then create, test, and deploy agents from your editor
Build AI tutors for your English teaching practice — long-form worked example for independent professionals deploying agents to clients (the same shape applies to consultants, fitness trainers, and other specialists)
Give your agent an API — register any HTTP API (AWS Lambda, Cloud Run, your own backend) as a named endpoint; the agent calls it and the key never leaves your secrets
Pin files to your agent's menu — put dashboards and key files one tap away in the chat header on desktop, web, and mobile
Connect a Supabase database to your agent — give the agent its own schema in a project you own; it builds tables by conversation, users read and write rows, only you change the structure
Create a self-hosted agent with an AI assistant — copy-paste prompt for Claude, Cursor, Windsurf, etc. that walks you through the full setup interactively

For owners who want to run the agent daemon themselves — configure Genie, choose tools, own the runtime:

Self-Hosted overview — when self-hosting makes sense, what's in this section
Runtime Modes — Mutiro-hosted, self-hosted with Genie, and brain swap
Configuration — .mutiro-agent.yaml reference
Swap the Brain — replace Genie with your own stdio bridge process
Brain Contract — the full chatbridge protocol reference for implementing a brain

References for anyone driving Mutiro from code or the terminal:

CLI reference — every mutiro command and flag
Agent daemon — what the host process does at runtime, message flow, history limits
pi-brain — minimum viable brain over the stdio bridge (start here when learning the protocol)
openclaw-brain — external brain implementation over the stdio bridge
claude-agent-brain — Anthropic Claude Code agent over the bridge