# Uniloom > Uniloom is one home for every AI coding tool on a Mac. It shows every AI coding session live, across 16 agents, and gives them one memory, one set of skills and rules (Toolkit) and one set of connections (Connect). Everything is stored locally in SQLite. Uniloom is also an MCP server: a connected agent reads and writes the workspace through it. How to use these docs: every page below is plain markdown at the link given. Fetch only the pages a question needs; `llms-full.txt` beside this file holds every page in one file. "MCP tool reference" lists the tools Uniloom gives an agent. The guide describes the app as shipped; where it says what Uniloom does not do, believe it. ## Get started - [Welcome to Uniloom](https://uniloom.dev/docs/index.md): Uniloom is one home for every AI coding tool you use: every session on your Mac live, one memory, one set of skills and rules, and one set of connections for 16 agents. - [Install and first ten minutes](https://uniloom.dev/docs/overview/getting-started.md): Install Uniloom, sign in, walk through the four setup steps, and get your first session and memory. - [Core concepts](https://uniloom.dev/docs/overview/concepts.md): The few ideas Uniloom is built on: sessions, agents, projects, memory, Toolkit and Connect, and workspaces. - [Uninstalling](https://uniloom.dev/docs/overview/uninstalling.md): Remove Uniloom from a Mac with one command. It takes Uniloom out of your agents and keeps your workspace unless you ask for it to go. ## Use Uniloom - [Sessions](https://uniloom.dev/docs/features/sessions.md): Sessions shows every AI coding session on this computer, live, so you see which one needs you, which is working and where each one runs. - [The map and the list](https://uniloom.dev/docs/features/sessions/list.md): How to read the Sessions map and list, what a row shows, the row menu, parking a worktree, and finished and past sessions. - [Start a session](https://uniloom.dev/docs/features/sessions/start.md): Start a session from the map or list, from New session, in any folder or from a doc, or open a Terminal, your own shell, the same ways. - [Your terminal and Uniloom](https://uniloom.dev/docs/features/sessions/take-over.md): Continue a session from your own terminal inside Uniloom, or hand a Claude Code session Uniloom runs back to your terminal. - [Worktrees](https://uniloom.dev/docs/features/sessions/worktrees.md): Every session on a git project starts in a new worktree of its own; what it carries over, the project's worktree commands, its port, Run preview with Pick, and how it is cleaned up. - [Workers](https://uniloom.dev/docs/features/sessions/workers.md): A session can start workers, agent runs of any agent in their own worktree, and wait for them; where they show, how you approve them, and the limits. - [Ask Uniloom](https://uniloom.dev/docs/features/sessions/ask-uniloom.md): Ask Uniloom is one box where you type what you want done; Uniloom starts a session on your own agent that finds the right project and starts the work there. - [Cloud sessions and phone](https://uniloom.dev/docs/features/sessions/cloud.md): Run a session in your agent's cloud so it keeps going while your computer sleeps, or continue a session running here from your phone. - [The terminal and Conversation](https://uniloom.dev/docs/features/sessions/terminal.md): A session Uniloom runs lives in a real terminal, with a Conversation view that reads it back as messages and tool steps. - [The side panel](https://uniloom.dev/docs/features/sessions/side-panel.md): The panel beside a session shows where it runs, what it made and how far it has got, what its branch changed, and every prompt it was given. - [Tabs, columns and names](https://uniloom.dev/docs/features/sessions/tabs.md): Open sessions as tabs, show up to four columns side by side, name a session, and how agents in one project learn about each other. - [Review the branch](https://uniloom.dev/docs/features/sessions/review.md): Review a finished branch line by line and merge, keep or discard it, and commit and sync a session's work from the Changes panel. - [Files, memories and editing](https://uniloom.dev/docs/features/sessions/files.md): Open a file beside a session, read it with blame, changes and history, see what memories say about its lines, comment for the agent, and edit it. - [Status and flow](https://uniloom.dev/docs/features/sessions/status.md): How Uniloom decides whether a session needs you, is working or is idle, what it cannot see, and how a session's flow shows which step of its work it is on. - [Home](https://uniloom.dev/docs/features/home.md): Home is where Uniloom opens, showing what needs you one card at a time, with a box to ask your agent anything. - [Ask from Home](https://uniloom.dev/docs/features/home/ask.md): The box on Home asks your own agent to find, read or change anything in the workspace, with files, skills and mentions, and can resume a past session. - [Approvals, Stop and agent differences](https://uniloom.dev/docs/features/home/approvals.md): How Home's agent asks before it acts, how autonomy and Stop work, and what differs between the agents. - [Tasks](https://uniloom.dev/docs/features/tasks.md): Tasks lists what waits on you (your to-dos, issues, reviews and messages) with today's meetings, and lets you run one of your automations on any of them. - [To-dos, reminders and meetings](https://uniloom.dev/docs/features/tasks/todos.md): Write your own to-dos in Tasks, set a reminder on a to-do or on any task, and see today's meetings above the Overview. - [Where tasks come from](https://uniloom.dev/docs/features/tasks/sources.md): How Tasks reads GitHub, your trackers, chat and mail, what happens when a tool is down or holds more than Uniloom reads, and how Messages stays in the open workspace. - [Choose what Tasks shows](https://uniloom.dev/docs/features/tasks/choose.md): Choose what each tool sends to Tasks: which git repositories, projects, channels or mail it looks in, which lists it reads, and your own searches. - [How Uniloom learns an app](https://uniloom.dev/docs/features/tasks/learning.md): How Uniloom learns to read a tracker it was not built for, the check that keeps every call read-only, and what Tasks keeps on this computer. - [Agent runs](https://uniloom.dev/docs/features/tasks/agent-runs.md): Start an agent run from a task, approve its plan and what it ships, let Autopilot run the steps, and finish, stop, resume or discard a run. - [Agent run rules](https://uniloom.dev/docs/features/tasks/agent-run-rules.md): The agent run rules that guard every agent run: what is allowed, what asks first, what is always refused, how to answer a held call, how to add your own rules, and what the rules cannot see. - [Which automation runs](https://uniloom.dev/docs/features/tasks/workflow.md): Which automation runs on which kind of task, and how a skill your project keeps can replace a starter's steps while the approvals stay. - [Receipts](https://uniloom.dev/docs/features/tasks/receipts.md): A receipt is what an agent session ran, checked against what it said: what it shows, what flags it, and how to copy or share it. - [Projects](https://uniloom.dev/docs/features/projects.md): Projects holds the docs written in each of your projects and their code: make and group projects, switch between Docs and Code, and narrow the list to one project. - [Write a doc](https://uniloom.dev/docs/features/projects/editor.md): How to write a doc in Uniloom: the editor's blocks and shortcuts, images, Figma links, diagrams, links between docs, and how saving works. - [The doc tree](https://uniloom.dev/docs/features/projects/tree.md): The Docs list in Projects: what opens when nothing is open, how to add a doc, what the tree shows, and how to reorder, move and group. - [Doc types](https://uniloom.dev/docs/features/projects/typed-docs.md): The five doc types (Plan, Design, Review, Bug and Note), the line each shows under its title, and how your agent saves a doc at a milestone. - [Files and changes](https://uniloom.dev/docs/features/projects/code.md): The Code tab of Projects: browse a project's files and worktrees, open a file with blame and changes, and commit from the Changes list. - [Group and project settings](https://uniloom.dev/docs/features/projects/settings.md): A group's or project's Settings: its folder, its agents, worktree commands, what done means for an agent run, hooks, the sources it works from, and what memory it keeps. - [Memory](https://uniloom.dev/docs/features/memory.md): Memory is what Uniloom remembers from your agents' work, shared by every agent you connect: how it fills, how it reaches a session, and where to read and fix it. - [The Memory screen](https://uniloom.dev/docs/features/memory/map.md): The Memory screen: a map of your memories on the files they are about, coloured by use, with a list to read, search, correct, pin, hide and retire them. - [Session capture](https://uniloom.dev/docs/features/memory/session-capture.md): How Uniloom keeps your agents' sessions: which agents it reads, what a captured session holds, and how to turn capture off for everything or for one project. - [What a session is given](https://uniloom.dev/docs/features/memory/session-start-brief.md): What an agent session is given from memory: the brief at its start, the card of a session started from a task, and the controls that decide which memories your agents get. - [Search and save](https://uniloom.dev/docs/features/memory/recall-remember.md): How an agent searches memory with recall, saves and corrects a memory with remember, gets a lesson on the tool call it is about, and tells Uniloom which memories helped. - [Kinds of memory](https://uniloom.dev/docs/features/memory/decisions-and-facts.md): The five kinds of memory, how a decision learns what came of it, how long a memory stays in use, and how to choose what a project keeps. - [Tidying memory](https://uniloom.dev/docs/features/memory/tidy-and-measure.md): How memory stays small and right: tidies your agent proposes, AI memory care each night, lessons from what you decline, lessons that become skills, and how to see whether memory helps. - [Activity](https://uniloom.dev/docs/features/activity.md): Activity shows your agent sessions over time: a summary of the range, cards that compare it with the days before, a board you can arrange, and where its numbers come from. - [Artifacts](https://uniloom.dev/docs/features/artifacts.md): Artifacts are the HTML pages and slide decks your agents publish for you to look at, kept in your workspace with every version. - [Widgets](https://uniloom.dev/docs/features/widgets.md): A widget is a small card your agent builds for Uniloom and shows on Home, Activity or a session's side panel. Make one, change it, and see what it can read and do. - [Place and size a widget](https://uniloom.dev/docs/features/widgets/placing.md): Place and size a widget on its tab: drag it, resize it, use its bar or the keyboard on the Preview in Widgets. - [Widget API](https://uniloom.dev/docs/features/widgets/api.md): The reference an agent builds a widget from: files, uniloom.json, every call with its permission, the hooks, the format helpers, the card parts, and the checks each version passes. - [Automations](https://uniloom.dev/docs/features/automations.md): Automations are processes you define once: a morning brief on a schedule, or a fix, review or reply that runs on an item from Tasks. Make one, choose what starts it, and see what it may do. - [The thread and reports](https://uniloom.dev/docs/features/automations/thread.md): An automation's thread of runs and reports: answer the last run, read its receipt, see what it knows, and the limits every background run keeps. - [Recent tasks](https://uniloom.dev/docs/features/automations/recent-tasks.md): Recent tasks, the first step of a scheduled automation: what it reads, how it checks your live tools, what its report shows, and the variables its prompt can use. - [Built-in automations](https://uniloom.dev/docs/features/automations/built-in.md): What Uniloom runs on its own, what each one does, what turning it off costs, and the switch for each, under Automations › Built-in. - [Toolkit](https://uniloom.dev/docs/features/toolkit.md): Toolkit shows everything a project's agents read, one tab per kind, and lets you add, edit, remove and share skills, rules, MCP servers and hooks across your agents. - [Adding items](https://uniloom.dev/docs/features/toolkit/adding.md): Add a skill, command, subagent, rule, output style, instructions file, MCP server, hook or plugin in Toolkit, and what each form writes. - [Edit and remove](https://uniloom.dev/docs/features/toolkit/edit-and-remove.md): Edit and remove Toolkit items, choose which agents read a .uniloom/ skill, and manage what your agents keep for every project under User. - [Usage, changes and sync](https://uniloom.dev/docs/features/toolkit/usage-and-sync.md): How Toolkit counts how often each item ran, the changes chip that leads to Code, and the sync pill that says whether every agent matches .uniloom/. - [Suggestions](https://uniloom.dev/docs/features/toolkit/suggestions.md): Toolkit's Suggestions tab: the five kinds of change Uniloom proposes to a project's AI setup, when each appears, and what its button does. - [Start using .uniloom](https://uniloom.dev/docs/features/toolkit/start-using-uniloom.md): Start using .uniloom in a project: what it creates, what it moves, and what it protects. - [Project environments](https://uniloom.dev/docs/features/toolkit/project-environments.md): Which account each agent runs under, per workspace and per group or project, so work stays on the work login, including sessions you start in a terminal. - [Connect](https://uniloom.dev/docs/features/connect.md): Connect is where a workspace reaches outside itself: the agents it runs on, and the apps it reads your work from. - [Import](https://uniloom.dev/docs/features/connect/import.md): Bring what you already have into Uniloom: your editors' projects and your docs into Projects, other tools' memories into Memory, and artifacts you made elsewhere into Artifacts. - [An agent's dialog](https://uniloom.dev/docs/features/connect/dialog.md): An agent's dialog in Connect: the MCP servers it reaches and how each stands, the profile it runs under, and its model. - [How Uniloom installs into an agent](https://uniloom.dev/docs/features/connect/how-uniloom-installs.md): What connecting an agent writes into its config, what you see in a session, what turning it off puts back, and how to check that capture works. - [Integrations](https://uniloom.dev/docs/features/connect/integrations.md): Connect an app such as GitHub, Jira or Notion to a workspace: how each one signs in, what each state means, and how your agents read it. - [Connected MCP tools](https://uniloom.dev/docs/features/connect/connected-tools.md): The MCP servers your agents reach from Uniloom: the ones each agent's own config declares, and the apps you connected in Uniloom. - [Workspaces](https://uniloom.dev/docs/features/workspaces.md): A workspace holds your projects, docs and memory in one file, and is shared or private as a whole. Switch between workspaces at the top of the rail. - [Working in a shared workspace](https://uniloom.dev/docs/features/workspaces/sharing.md): Inside a shared workspace everything is shared. How to bring work in, why a teammate's doc is read-only, and how docs, memory, images and artifacts travel. - [Workspaces and invites](https://uniloom.dev/docs/features/workspaces/workspaces-and-invites.md): Create a workspace, invite people, what each role may do, and how to delete a workspace forever or remove a person. - [Search and command palette](https://uniloom.dev/docs/features/search.md): Press ⌘K to run a command, jump to a doc or session, or search your workspace. Also how the rail, the sidebar, tabs, and Back and Forward work. - [Trash](https://uniloom.dev/docs/features/trash.md): Docs and widgets you move to Trash stay there until you restore them or delete them forever. - [Themes](https://uniloom.dev/docs/features/theming.md): Uniloom's two themes, Mint and Forest, the nine workspace colours, and how to make the whole app bigger or smaller. ## Reference - [Uniloom as an MCP server](https://uniloom.dev/docs/features/mcp-server.md): Uniloom is an MCP server: every connected agent reads and writes the same workspace and memory, under one recall and remember contract. - [Connecting your agents](https://uniloom.dev/docs/features/mcp-server/connecting.md): Connect any of the 16 agents to Uniloom's MCP server: from Connect, with one click in VS Code or Cursor, or by hand. - [MCP tool reference](https://uniloom.dev/docs/features/mcp-server/tools.md): Every tool Uniloom's MCP server exposes, with arguments and defaults: docs, memory, workspaces, activity, your Tasks list, connected apps, other agents, artifacts, widgets, setup, automations, agent runs and capability discovery. - [Tool rules](https://uniloom.dev/docs/features/mcp-server/tool-rules.md): The rules every Uniloom MCP tool call follows: which workspace a call reaches, which secrets are removed, what an agent cannot do, and how a doc target is found. - [Curated MCP commands](https://uniloom.dev/docs/features/mcp-server/commands.md): How Home runs well on a small model: Uniloom picks one of its own tools from your words, fills in what it already knows, and the model fills in only the rest. - [Agents](https://uniloom.dev/docs/features/agents.md): The 16 agents Uniloom works with. Uniloom runs the command-line program you already installed and signed in to, so there is no API key to enter. - [Features by agent](https://uniloom.dev/docs/features/agents/features-by-agent.md): Which of the 16 agents has which feature in Uniloom, and why some cells are "no". - [Settings](https://uniloom.dev/docs/features/settings.md): What each Settings tab holds: Account, Workspaces, Appearance, Keyboard and About. Every tab is about this computer. - [Keyboard shortcuts](https://uniloom.dev/docs/features/shortcuts.md): Every keyboard shortcut in Uniloom, grouped by where it works, and how to change one. ## Privacy & trust - [Privacy and security](https://uniloom.dev/docs/features/privacy-and-security.md): Your workspace stays on your Mac. Here is everything that can leave it: your sign-in, four standing exceptions, a problem report you send, and requests and connections you start yourself. - [Where each piece of data lives](https://uniloom.dev/docs/features/privacy-and-security/where-data-lives.md): Where each kind of data Uniloom keeps is stored on your Mac, which of it a shared workspace can send, and which credential files Uniloom writes. - [Workspace sharing, precisely](https://uniloom.dev/docs/features/privacy-and-security/sharing.md): Exactly what a shared workspace sends and what it never sends, what a widget or live artifact can reach, and how to take back what you shared. - [Security details](https://uniloom.dev/docs/features/privacy-and-security/security-details.md): How Uniloom protects its local server, your agents' setup files and your workspace files, what it does not encrypt, and how IT teams manage it. - [Your data and export](https://uniloom.dev/docs/features/data-and-export.md): Each workspace is one SQLite file on your Mac. Where it lives, how to move it, export it, shrink it and restore it, and what an export leaves out. - [Licensing and updates](https://uniloom.dev/docs/features/licensing.md): Uniloom is free for individuals and Team is coming. How updates download, when they install, and what safe mode does. - [Platform and availability](https://uniloom.dev/docs/features/platform.md): Uniloom runs on macOS 13.5 or later, with one app for Apple Silicon and one for Intel. Linux and Windows are paused. - [What Uniloom costs your machine](https://uniloom.dev/docs/features/footprint.md): The most CPU, memory and disk Uniloom may use on your Mac, and how to check it yourself. ## Legal - [Terms of use](https://uniloom.dev/docs/legal/terms-of-use.md): The terms that govern your use of the Uniloom desktop application, including the license, acceptable use, warranty disclaimer, and third-party tools. - [Privacy policy](https://uniloom.dev/docs/legal/privacy-policy.md): How Uniloom handles your data: your work stays in local SQLite files on your device unless you deliberately share a workspace. Your account holds your name, email and plan.