Learn Mosaic

The choice is for this chat Turn on logging when you want this thread's work remembered in its pod. Nothing is posted when you choose. Not this thread keeps a shared-pod chat out of Activity. If an unselected 1:1 DM offers a pod, Not now keeps its personal-pod default; in a group DM it leaves the thread unlogged. You can change the choice later. A summary when the chat goes quiet Mosaic summarises a conversation after it goes quiet when Activity logging is on. The summary can include decisions, work done, work in progress and follow-ups. It is a summary of the chat, including earlier messages, not a copy of every reply. Where it goes A chat acting in a shared pod logs to that pod's Activity. An unassigned 1:1 direct message or a web chat acting across the whole organisation logs to your own personal pod. In Slack, you can choose a shared pod for a direct or group DM; a group DM can use only a pod everyone there may read. Not now keeps a 1:1 DM on its personal default and leaves a group DM unlogged. Activity readers can see the summary only where they have access. Turn logging off Use the chat's Activity logging control to choose Not this thread. Later messages stay out of Activity. A summary already published remains in its pod until someone removes it there. Ask the brain Answers draw on Open pods and the organisation. Choose a Private pod in the conversation when the answer should draw on it. To ask a question and get an answer drawn from the brain, ask your coworker in Slack or any connected AI tool. Browse a pod's documents To find a document by eye instead of asking, open a pod and browse its Knowledge tab. Read the live version, with its trust signals Every document shows owners, when it was last live, and a chip on each fact - Verified, Stated, or Inferred - and the agent's own reply grounds itself in what changed most recently, so it never contradicts a decision made minutes ago. Open any document from its pod to see the live version and who stands behind it. See exactly which facts an answer used To check which brain facts backed a reply, open the numbered Sources line underneath it. Connect your tools Connections act as you, never more - each one runs at your access, not the whole organisation's. Connect an app yourself, with its own nickname if you use more than one account. Approve what the agent does in your tools Set an outside-app action to run automatically, ask first, or stay off, down to the individual action. Capture what you learn, without a context switch Say "capture this" in Slack, or save a web chat answer; the draft waits in the review queue for your pod lead's decision. What you can ask / how to query the brain The 13 Core homes and how to choose one Propose a correction Open any document you can edit and use Suggest an edit - nothing changes until a pod lead approves it. Track your proposals See a proposal's status and, if it was rejected, the reason. Work the review queue A pod lead other than the author approves every durable change - the agent follows the same rule and can never self-approve. Approve, reject with a reason, or see the before-and-after diff before you decide. Confirm or flag a stale fact Confirm a stale-flagged fact is still true, or send it back for an update, in one click. Resolve facts that disagree When two sources disagree on the same fact, resolve the cluster at once instead of hunting each one down. See what depends on a document before it changes Check what else in the brain relies on a document before you change or retire it. Retire a document that's no longer needed Retire a document once nothing still depends on it - it stays in the audit trail, never quietly deleted. See your brain's health in one place See what's waiting on you, what's healthy, and what the Brain architect suggests, without piecing it together from separate pages. Feed your pod from where work happens Sources are what Mosaic reads; the brain is what it knows - a source is always a pointer back to the original, never a copy. The Slack channel directory keeps a channel's conversation pod separate from its single source-pod binding. Connect a folder, channel, or project so its activity feeds your pod. Fold a source into Activity right now Spot a catalogued source document that matters right now? Curate it straight from the Live Activity Stream row, instead of waiting on the nightly pull. Stand up a new pod Create a pod and seed its first documents. Tag what the brain holds Signal's own fields - kind, tier, freshness - are set automatically from how something reads. Tags are the one part that's yours: the shared words you and your coworker choose together to classify and find things. Let the brain grow and mend itself After onboarding, Activity Integration reads bound sources, brings new material into Activity, proposes new knowledge or changes to the facts it touched, and runs light upkeep. Maintenance keeps existing knowledge correct: its light pass runs with Activity Integration and its deep pass runs weekly. Enhance is the Brain Architect reflecting across the brain and proposing larger content or structure improvements. Review what Activity Integration proposed before it becomes truth. See the Brain architect's last run and its findings. Turn any of it down, or off Proactive AI is on for everyone by default and shown so you know it happens - Activity Integration, suggestions, and the introductions and digests it sends on its own. See everything Mosaic does without being asked, grouped by what it's for, and turn off exactly the part you don't want. Rate an answer or a suggestion React right where something was delivered - an answer, a skill run, a suggestion - never a separate page. Yes/Partly/No is required; Partly or No also opens an optional pick from six issue reasons (incorrect or unsupported, missing important information, didn't follow the request, hard to use, an action or tool that failed, or other) plus an optional note in your own words. Saving shows an immediate confirmation with an Undo right there. In Slack, a thumbs up or thumbs down emoji on any of your coworker's answers is feedback too: a thumbs up saves Yes and a thumbs down saves No. The first time you react to an answer, a short follow-up only you can see may appear in the thread, to add a note, say it was only partly right, or pick what went wrong. It is optional: ignore it and your reaction still counts. Feedback shows which answers work and which to improve. There's no feedback history or inbox for other members, Mosaic staff, or an ordinary MCP tool to browse. The Brain architect may read scoped feedback - rating, verdict, category - as evidence for improving pods you lead; your comment is included only if you agreed to share it. Nothing open to react to - an older answer, a general impression, or something that happened outside Mosaic? Ask your coworker to give feedback about anything. Accept a suggestion, or dismiss it with a reason: not relevant, already handled, not now, wrong, too often, or your own words. See where Mosaic reaches you, and how often Choose Slack, a channel, or email, and how often, instead of remembering to check a queue. See exactly what Mosaic sent you and whether it reached you, under What Mosaic sent. See what skills exist and how to ask Kinds, and who writes each: Mosaic maintains the included actions that only it can create, while your organisation authors and adapts its own reusable skills. You run one by asking, in Slack or a connected AI tool - there is no run-it-now button here. Browse the catalogue of included actions, prompt skills, and skill chains, with the exact words to ask for each. Write or improve a skill A skill chain runs several of your own skills in order, handing each result to the next. Pick a specific model on the skill's own form when its job is worth extra reasoning power, or leave it blank for your coworker's usual default; limit which tools it can reach the same way. A skill can also be a short script you write yourself, which runs exactly as written once an org admin has reviewed and pinned it. Every skill ends with its technical details: the tools it uses in order, the data it reads, related skills, known gotchas and when it is done, so a run starts with what it needs. Draft, check the wording, and publish your own skill. Take a fresh update to an adopted skill, keep your own version, or merge the two - never a silent overwrite. Check a skill's quality before you trust it See how each skill's results land: how often people say it helped, and which issues come up most. Turn a repeatable job into an automation, and try it first Approve once, then it runs on its own - and you can trial a skill against real conditions before you commit to it. Turn a repeatable job it noticed in your channels into an automation, or set one up yourself. Put work on a recurring schedule Hand a skill to the agent on a schedule - create, edit, pause, or cancel it whenever you like. Onboard someone with the right role Invite someone with a role from day one, or grant a whole team access at once instead of person by person. Decide what applies without a person, per pod Decide whether changes to your brain apply on their own (Automatic, the default) or wait for a pod lead (Gated), for the organisation and per pod, and how you hear about them. See what Mosaic approved on its own See what changed, who proposed it, who chose the review setting that allowed it, and undo it in one place. See patterns worth every pod knowing Every pod's coworker reads Org Signal - patterns that held in more than one pod, or that a lead promoted deliberately. Choose a model without being an expert Compare what each model is good at and whose terms govern your content, before you pick one. Prove the brain is controlled Every change to the brain - who proposed it, who approved it, and when - in one plain-sentence, tamper-evident log. See what your coworker costs Included actions are marked as using no allowance. If your organisation has billing access, Billing shows the balance, plan, and spend. See the balance, this period's usage, and who spent what. See capture rate, active members, and how well runs are succeeding, not just what they cost.