Agents

Kuuumba agents live inside your workspace and work from unified memory, so one prompt can reach your email, tasks, docs, calls, channels, and connected tools. They can search and read anything you can see, draft emails and documents, create and update tasks, post in channels, and run on a schedule. Everything they do is attributed to you and bounded by your permissions.

To start a new chat use shortcut c + a

The Agents tab

The Agents module lists every chat you've had, named automatically from your first message, and you can pick any of them back up where you left off. Ask for summaries, clarifications, drafts, or a status check. Agents answer using everything in your workspace, not just the current thread.

Use shortcut g + a to go to Agents, and j and k to move up and down through chats.

Full workspace context

Agents answer from unified memory, a continuously refreshed view of your email, messages, tasks, docs, calls, and connected tools, not just your chat history. That's why you can ask a broad question ("what's the latest on the Acme deal?") without pointing them at anything.

When you want to be exact, @mention a specific channel, doc, task, or thread right in your prompt. That pins the context and skips the search, so agents work from precisely the item you mean. See Mentions in Agents.

What agents can do

Agents have a broad tool set and will reach for the right tools on their own. In one conversation they can:

• Search & read — find and read across docs, threads, chats, and records
• Write docs — create, edit, and rename documents
• Tasks & CRM — create tasks and set status, priority, assignees, and other properties on tasks and company records
• Email — draft a reply in an editable composer for you to review and send
• Channels — read a channel or thread and post messages back into it
• Calls — read call recordings and transcripts and turn them into tasks or summaries
• Web — search the web and fetch pages when the answer isn't in your workspace
• Import — pull content in from Notion, Linear, and Slack when you migrate

For copy-paste prompts that put these to work, see Agent recipes.

Agents in channels

Mention @Kuuumba in any channel to pull agents into the conversation with the whole channel as context. They show up in the normal @-mention typeahead, read the thread, and reply inline. Agents can also post their results back into a channel or a specific thread as part of a longer task.

Automations

Agents can run on a schedule. An automation is a saved prompt that runs on its own, like a daily inbox brief, a Friday team recap, or a weekly reminder to triage stale tasks.

To create an automation

• Go to the Agents Block, press Agents in the top right corner, then Automations.
• Fill out the form with your instructions and the schedule.
• Hit Create and wait for your automation to run.

Choosing a model

A model picker sits in the chat input. You can choose a model per chat depending on how difficult you deem the question. Your choice is remembered for that chat and carried forward when you start a new one from the composer.

All supported models are available from the picker, so choose the one that fits the task and response time you want.

MCP setup

Kuuumba's MCP server lets AI clients like Claude Code, Codex, and Cursor work with your workspace directly. Connect once by going to Settings → MCP server and authorizing with your Kuuumba account, or point a client at the endpoint below.

https://mcp-server.kuumba.dev/mcp

See MCP setup for client-by-client instructions and the full MCP tool reference.

@mentioning

Agent chats are markdown entities like everything else in Kuuumba, so you can @mention a chat from a doc, task, email, or channel to link back to a conversation. And it works the other way too: @mention docs, tasks, channels, people, and more inside a prompt to hand agents exactly the context you want.