Back to docs

Platform Overview

title: Platform Overview

--- title: Platform Overview description: High level overview of Ekairos platform layers and goals. ---

Platform Overview

Ekairos is a platform for building domain aware AI systems. The platform is composed of three layers:

1) Ekairos Workspace (monorepo of packages) 2) Ekairos Core (this app, docs, demos, internal UI) 3) Domain apps (example: esolbay-platform)

The Workspace provides the core primitives:

  • Domain modeling and schema composition.
  • Durable threads and workflows.
  • Runtime bootstrapping for Next.js and serverless.
  • Dataset tooling for long running data preparation.
  • Sandbox execution for agentic code.
  • Testing tools for long running flows.

Ekairos Core is the reference host that ties these pieces together for internal use. The Core app is also a demo surface to validate the APIs, the UX for threads, and the integration strategy for domain apps.

Architectural goals

  • Single root domain per app, with composable subdomains.
  • Runtime configuration driven by a bootstrap module (example: src/ekairos.ts).
  • AI agents that can reason over domain context and execute actions safely.
  • Long running workflows (use workflow) that can run in production and testing.
  • Clear separation between schema (structural) and DOMAIN.md (narrative).

Key terms

  • Domain: the schema and actions for a bounded area of the system.
  • Thread: an AI loop defined by actions and narrative, executed by a workflow runner.
  • Runtime: the environment bindings that connect threads to databases and services.
  • Dataset: a persistent, auditable artifact of data preparation.
  • Sandbox: isolated execution environment for agent tools.

Primary entrypoints

  • Domain DSL: @ekairos/domain
  • Thread engine: @ekairos/thread
  • Runtime integration: @ekairos/domain/runtime (runtime + domain/MCP) + @ekairos/thread/runtime (thread execution)
  • Next integration: @ekairos/domain/next (withRuntime)