Contorium is the runtime continuity layer for AI coding agents.

It keeps AI continuously aligned with your workspace across sessions, tools, and models.

AI coding tools reset context. Contorium maintains runtime state.

.contora / runtime state

We define a new layer in AI coding systems.

Runtime Continuity Layer

Not memory. Not retrieval. Not context injection.
A persistent runtime state layer between AI agents and your workspace.

Why this category didn't exist before

Not a feature comparison — a paradigm shift.

The old world

  • Memory tools store snapshots
  • RAG systems retrieve context
  • Agents restart every session

Contorium

Maintains a continuous runtime state of your workspace

Current Focus

The active intent of your work.

AI always knows what you're building right now.

Workspace State

The real-time structure of your project.

Files, changes, and working context.

Session Continuity

Work continues across sessions without reset.

No re-explaining. No context loss.

Before
  • AI forgets your project every session
  • You rebuild context repeatedly
After
  • AI continues your workspace state across sessions
  • No reset. No re-explanation.

Not a tool. Not a plugin. A runtime layer.

Contorium introduces runtime continuity as a foundational layer for AI coding systems.

Integrations

Works with AI coding agents via MCP and native extensions.

Cursor VS Code Claude Code Codex MCP-compatible tools

Install Contorium

IDE extensions for runtime tracking in the editor. MCP runtime for Codex, Claude Code, and agents.

MCP runtime

Codex & Claude Code

Agent-callable tools—get_workspace_context, store_memory, and more. Complements the VSIX extension for Codex, Claude Code, and Cursor Agent workflows.

View MCP setup
Quick setup
npm run build:mcp

# Codex
codex mcp add contorium \
  -- node ./bin/contorium-mcp-launch.cjs

# Claude Code
claude --plugin-dir .

Codex guide → · Claude Code → · docs/MCP.md

Contorium is a runtime continuity layer that keeps AI coding agents aligned with your workspace across sessions and tools.