Quick Start

The fastest path into Albumi is to let an AI agent do the entry work. You bring existing documentation — a wiki page, a README, architecture notes, free-form descriptions — and the agent turns it into applications, integrations, data objects, and business capabilities. You review the result as an Architecture Change Request and approve what you want to keep.

If you would rather enter data by hand from the start, follow Manual Start. The rest of this page assumes the AI path.

1. Sign up

Go to my.albumi.app and sign in with Google or Microsoft. You get a workspace immediately — no approval queue. Name it something your organization will recognize; you can rename later.

2. Connect your AI agent

Albumi exposes itself to AI agents as an MCP (Model Context Protocol) server. It runs on your machine, authenticates to your workspace, and gives the agent four tools: pull, validate, audit, and push.

Install the CLI (required, all clients)

npm install -g @albumi/cli
albumi login

A browser opens, you confirm, and a token is saved to ~/.albumi/credentials.json. The albumi binary now works both as a command-line tool and as the MCP server process your agent will talk to.

Register the MCP server with your agent

Claude Code — plugin (one step, also installs a routing skill):

/plugin marketplace add albumi-eam/albumi-cli
/plugin install albumi@albumi-cli
/reload-plugins

/reload-plugins activates the MCP server and skill for your current session. MCP tools (pull_workspace, validate, audit, push_workspace) are then always available to the agent. The routing skill auto-activates when your message contains Albumi-specific context — for reliable activation, invoke it explicitly with /albumi-workspace at the start of an Albumi-related session.

Or manually register without the plugin/skill:

claude mcp add albumi -- albumi

Cursor, Claude Desktop, other MCP clients — see MCP Server Setup for the exact per-client configuration. Node.js 20+ is the only prerequisite.

3. Generate a landscape from your documents

Open a conversation with your agent and point it at whatever you have. Examples that work:

“Here is our architecture wiki page. Create a complete Albumi workspace — applications, integrations, data objects, and business capabilities. Then run validate and audit.”

“Here is a Confluence export of our integration catalogue. Model the source systems, the targets, and the data objects. Infer the business criticality where it is described.”

“We run Salesforce (CRM), SAP (ERP), and a custom data warehouse. Salesforce sends customer data to SAP daily via REST. SAP sends financial data to the warehouse nightly by batch file. Model this.”

The agent produces a workspace with valid identifiers, referential integrity, and all required fields. The more specific your input, the better the result — see Generate Architecture for what makes a good source document.

4. Review the Architecture Change Request

The agent does not write directly into your landscape. It opens a pull request against your architecture — an Architecture Change Request (ACR) — and pushes the generated content there. You review the ACR in the UI, adjust anything the agent got wrong, approve, and implement.

See Governance Overview for how ACRs work end-to-end. The short version: nothing enters your landscape until you approve it.

5. What to do next

With a first draft in place:

• The Edit Loop — the full pull → edit → validate → audit → push cycle for making ongoing changes through the agent, with real prompts and examples.
• Key Concepts — the five building blocks of Albumi, if you have not read them yet.
• Entities Overview — how the seven entity types relate and which to reach for when the choice is not obvious.
• AI Use Cases — prompt patterns for incremental updates, impact briefings, audits, and migration planning.
• Manual Start — if you want to continue without the agent.