Core Concepts

A few concepts explain almost everything in Wato. Learn these and the rest of the product follows.

Workspace

A workspace is the top-level container — the same thing as your organization. It holds your teams, members, billing, and settings. You become org admin of the workspace you create. Most things in Wato live inside a team, not directly in the workspace.

Team

A team is the main boundary for everything an agent touches: memory, connectors, skills, agents, automations, traces, and review. A workspace can have many teams (Sales, Engineering, Support…), and a user only sees a team's private content once they're a member. Team admins manage membership and settings. Your workspace starts with a default team called General.

Each MCP connection runs in one team's context — that's how Wato scopes what an agent can see and do.

Memory

Memory is your team's durable, reviewed knowledge — Markdown pages an agent can read across sessions. Changes are drafted, submitted for review, and approved before they become trusted memory, so nothing unreviewed leaks into agent context. See Getting Started and Common Workflows.

Connectors & tools

A connector is an integration (GitHub, Linear, Notion, a CRM, a custom MCP server) that exposes tools an agent can call. Admins approve connectors per team and can allow or block individual tools, so a team gets exactly the access it should — no more. Agents call approved tools through the Wato MCP gateway, never with raw credentials.

Skills

A skill is a reusable, versioned unit of capability — a SKILL.md plus optional bundled files — that captures "how we do X." Agents discover and load a team's published skills on demand. See Skills.

Agents

An agent is an AI assistant working inside a team's context. It connects through the Wato MCP gateway from a client like Claude Code, Codex, or Cursor (see Connecting AI Clients), or runs in Wato's own agent workspace. Whichever it is, the agent only gets the team's reviewed memory, approved tools, and published skills — and its activity is recorded as traces.

Automations

An automation is a workflow that runs on a trigger (a schedule, a webhook, or a connector event) and carries out steps — calling approved connector tools and asking an agent to make decisions. Build one on a visual canvas, or turn a chat with a connected agent into one. See Quickstart: Your First Automation and Build Automations from a Chat.

Artifacts & dashboards

An artifact is a published, access-controlled web page (an app, report, or dashboard). A live dashboard is an artifact that pulls fresh data at runtime through approved connectors, with credentials kept server-side. See Artifacts & Dashboards.

Review & proposals

Memory changes and skills don't become trusted the moment they're written. They move through a review flow — a draft is submitted as a proposal, inspected, and merged/published. This is what makes agent-generated content safe to rely on.

Traces

A trace records what happened on an MCP call — which tool, who called it, which team, the input, the result, and when. Traces give you audit, debugging, and compliance across everything agents do.

How it fits together

Workspace (organization)
└─ Team
   ├─ Memory        reviewed knowledge
   ├─ Connectors    approved integrations → tools
   ├─ Skills        reusable, versioned capabilities
   ├─ Agents        AI clients working in the team's context
   ├─ Automations   triggered workflows over tools + agents
   ├─ Artifacts     published pages & live dashboards
   └─ Traces        audit of every MCP call