EmbrAIon
Where sparks become AI-built products
A portable engineering system for making AI-assisted software development structured, reviewable, reusable, and trustworthy.
System
Engineering structure around the model
EmbrAIon sits above individual languages and frameworks. A Unity project, Python service, web application, or another repository can share the same Core while keeping project-specific knowledge and rules in its own overlay.
The framework stays model-agnostic: hosts own their available models and defaults, while project configuration can add explicit routing overrides without weakening privacy, validation, or review policy.
Capabilities
Independent dimensions, composed when needed
Agents
Explicit responsibility across lead, worker, reviewer, architect, analyst, validator, researcher, and steward roles.
Skills
Reusable procedures for repeatable classes of engineering work.
Rules
Required, forbidden, protected, and security-sensitive behavior expressed as policy.
Workflows
Composable execution order for substantial engineering tasks.
Routing
Complexity, access, privacy, and host-aware task routing without hard-coding a universal model catalog.
Adapters
Host projections for Codex, GitHub Copilot, Claude Code, providers, and portable bundles.
Tools
Deterministic validation, security, diagnostics, synchronization, and worktree behavior.
Evals
Checks that AI behavior follows the intended engineering contract.
Task flow
From goal to verified delivery
Distribution
Install once. Opt in per repository.
EmbrAIon is distributed as a Python CLI. Projects explicitly opt in and keep their own versioned overlay.
$ pipx install embraion
$ embraion init
$ embraion install --host codex --destination .Built by GORYNED