Skip to content
AggregaAggrega
Select language: English

Aggrega

What is Aggrega

Aggrega is an open-source compiler for AI-era architecture. Describe the components of an application and how they connect; Aggrega turns them, deterministically, into the deployment files that run it.

  • Use the languages you know. A compiler for the languages you already use: YAML in; Helm charts, Kubernetes manifests and Compose, Terraform or CloudFormation files out.
  • Developers describe, Ops implement. Developers declare components, their features and who uses whom; the platform team decides how and where each one runs.
  • One application, many platforms. The same application runs on Docker Compose on a laptop and on Kubernetes in production: changing target changes a mapping, not your code.
  • Reproducible and governed. Same input, same output, with every dependency pinned in a lockfile. Your policies plug into the compiler and stop a build that breaks them, saying what to fix.

Aggrega is Italian for ‘it brings together’: the name says what the tool does, gathering the parts of an application into one description.

  • Use it with your AI assistant

    Paste one prompt into an assistant that can run commands: it installs Aggrega, adds the Aggrega skill and connects to the Aggrega MCP server, checking with you before it installs anything.

  • Quickstart

    Install Aggrega, write a runnable example into a folder and run it on your machine, then grow it one step at a time: an AI agent on a local model, or a web app with PostgreSQL.

  • Learn Aggrega

    A path through the documentation in reading order, from what Aggrega is for to the language reference.

Install

On Linux and macOS:

Terminal window
curl -fsSL https://aggrega.ai/install.sh | sh

On Windows, in PowerShell:

Terminal window
irm https://aggrega.ai/install.ps1 | iex

Either command installs the aggrega binary for the current user. Install covers what the command does, Homebrew, Scoop, winget and the manual download; Connect the MCP server connects an AI assistant. The Aggrega extension for Visual Studio Code brings validation, navigation, hover and completion to Aggrega YAML, backed by the aggrega binary.

The documentation

User-facing documentation for the Aggrega language and compiler, written in English and translated into every other language the site offers. It has three parts.

  • Architecture — the “why” of Aggrega. Vision, positioning, mental model, compiler pipeline at conceptual level, enterprise governance, catalog, extensibility, design choices. Follow the numbered chapters 01 through 16 in order for first-time reading.
  • Authoring Guide — the “how to sit down and author” of Aggrega. The authoring loop, the four collaborating roles, the kind-to-role map, and how to connect an AI assistant through the Model Context Protocol server. Start with Authoring Aggrega files; to have your AI assistant install and set up Aggrega itself, paste the prompt in Set up Aggrega with your AI assistant.
  • Language Reference — the “what” and “how to use” of Aggrega. Start with Status of the Language for the current state of readiness, then follow the numbered chapters 01 through 09 in order for first-time reading. Unnumbered subdirectories (entities/, vocabulary/, templates/, providers/, plugins/, schema/, cli/, errors/, examples/) are reference catalogs — jump to specific items as needed.

About Aggrega

The Aggrega compiler is open source under Apache-2.0, and stays that way: there will be no closed edition of the compiler, no open-core. All the features of the compiler will be available in the open source version.

The ecosystem is built on four pillars:

  1. The Separation Model (OAM-Inspired): Design guidance that separates Developer Intent (the abstract “What” and “How much”) from Platform Ops physical reality (the physical “How” and “Where”). The compiler does not enforce the split: it accepts any authoring pattern that passes schema validation.
  2. The Semantic DSL & Operational Ontology: Developers map their architecture using a closed set of universal cloud primitives (e.g., server, worker, stateful-server) on Aggrega’s reserved component.aggrega.ai/type label. All advanced context—such as AI agent autonomy, generative capabilities, or data compliance—is layered on top via static Duck-Typing labels, and a catalog classifying components on an axis of its own adds a label key of its own.
  3. Programmable Policies & Validation: The core Go compiler is unopinionated. Corporate governance is driven entirely by industry standards and pluggable WebAssembly: JSON Schema 2020-12 (authored via a compact YAML-shaped shorthand or full JSON Schema YAML) for catalog validation, the governance socket’s label-transform role for dynamic topology tagging, and its verdict roles for evaluating architectural security boundaries.
  4. The AOT Compiler & Execution Handoff: A strict compilation pipeline translates the DSL into a capability-agnostic Intermediate Representation (IR) graph. The IR captures the topology of components and the uses edges between them (who uses whom); the compiler passes this graph to the WebAssembly (Wasm) governance socket to detect transitive trust vulnerabilities, then executes Wasm rendering plugins to generate the precise proxy and guardrail configurations. Aggrega writes these artifacts to disk and hands them off to other tools, like CI/CD pipelines or runtime orchestrators, which deploy them.

License

The Aggrega documentation is released under the Apache License, Version 2.0; the full text is in the LICENSE file of its repository.