Resources and actions
Typed, introspectable create, read, update, destroy and generic actions are the core abstraction. Everything else builds on them.
Home / Open frameworks / Ash Framework
Open framework · Elixir · MIT
Define resources and actions. Generate the rest from them.
Ash is a declarative, resource-oriented application framework for Elixir. You describe resources, their actions and their policies; extensions derive the data layer, GraphQL and JSON:API endpoints, admin screens, state machines and LLM tools from the same definition. It runs on the BEAM, the Erlang virtual machine built for concurrent, fault-tolerant systems.
Core concepts
How Ash Framework is put together.
Our assessment
Strengths and trade-offs.
Our view as of November 2026, written for teams choosing a foundation for business software.
Strengths
- One declaration drives the schema, APIs, authorisation, jobs and state, with less duplicated code.
- Policies are enforced at the action layer, so every API and interface applies the same rules.
- The BEAM runtime: very high concurrency, fault tolerance and real-time features.
- Many mistakes become compile-time errors rather than production failures.
- A built-in path to AI agents: actions as LLM tools and MCP servers.
- Mature (3.x), MIT-licensed, with an active community and commercial support.
Trade-offs
- A smaller talent pool: Elixir is less common than Python, and Ash's own docs call it "still niche".
- A steep learning curve: Elixir, functional programming and Ash's declarative style at once.
- No ready ERP or CRM application layer: business applications are built, not installed.
- AshAdmin is basic next to Frappe's Desk; business users get less self-service configuration.
- More than a simple CRUD application needs.
Fit
What it suits. What it doesn't.
Well suited to
- Custom, domain-rich products and SaaS platforms
- Real-time, high-concurrency systems: IoT, telemetry, marketplaces, collaboration
- API-first platforms that need GraphQL or JSON:API from one model
- Exposing a business domain to AI agents through MCP
Less suited to
- Off-the-shelf ERP, accounting or HR needs
- Teams without Elixir capacity or a plan to build it
- Small internal CRUD tools
How we work with it: We assess where Ash fits, typically real-time or API-first services around an ERP, and can prototype them with ERPNext as the system of record.
Compare
Full comparison →The other open frameworks.
See all three side by side, with our guidance on when to choose each.
Frappe Framework
The metadata-first base for ERPNext, where records, permissions, workflows and reports matter.
Read the guide →
Framework M
An open, metadata-driven Python framework whose schemas agents can read directly.
Read the guide →SourcesAsh FrameworkWhat is Ash? (documentation)Ash repositoryAsh ecosystem