# Semryn — Your codebase, understood

> Semantic software infrastructure for AI coding agents.

Semryn is the product/platform, powered by AIR — AI Intermediate Representation. The planned Semryn Context Engine should give coding agents a semantic understanding of existing repositories: relevant symbols, dependencies, types, effects, source references and task constraints.

## Value proposition

Less repeated search and rereading; task-specific context; smaller reviewable changes; validation before application. Token, repair-cycle and total-cost improvements must be measured in repository pilots. No universal productivity multiplier is proven.

## Existing workflows

Customers should retain their source languages and existing repositories. No application rewrite in AIR is required. Intended languages include TypeScript, JavaScript, PHP, Python, Java, Rust and C#. Planned agent integrations include Codex, Claude Code, Copilot, Gemini and custom agents. These are not delivered integrations or vendor partnerships.

## Current status

Semryn’s current foundation is the working AIR local graph-engine prototype, version 0.0.3. It implements binary native/web semantic graphs, typed construction, validation, revisioned attribute edits, local binary control, native/static HTML backends and host-owned observable checks.

Existing-repository indexing, task-specific repository context, source-language adapters and semantic source patches are future work. The interactive map on the website is an illustration, not a live repository indexer.

## Next milestone

A read-only context pilot for one source language and one agent adapter, with task instrumentation. Then controlled source patches checked by parser, build and task validation.

## Resources

- [Product strategy](strategy.md)
- [Capabilities for agents](for-agents.md)
- [Structured metadata](air.json)