One of the biggest problems in AI-assisted development isn’t code generation.
It’s continuity.
Every time we:
- open a new chat
- switch models
- move between Claude, Codex, Cursor, or Gemini
we risk losing project understanding.
Contorium was originally built as a persistent project memory layer.
Today’s update takes that idea further.
Runtime Dashboard
We’ve introduced a new Runtime Dashboard designed to keep project awareness visible without becoming distracting.
Passive Mode
A lightweight status line showing:
- current task
- latest change
- active agent
Expanded Mode
A deeper runtime view including:
- Impact Graph
- Function Call Chains
- Runtime Handoff
- Copy To AI
The goal is simple:
make project state visible while keeping developers focused on coding.
⸻
Semi-Auto Runtime Injection
One challenge with AI handoffs is control.
Automatic prompt injection can feel intrusive.
Manual copy-paste creates friction.
Contorium now introduces a middle ground.
When a new AI conversation starts and an active runtime exists:
[?] Runtime active.
Inject current project state? (Y/n)
Developers choose.
This preserves continuity without taking control away from the user.
⸻
MCP Standardization
We’ve continued moving Contorium toward a fully standardized MCP architecture.
The goal is simple:
Configure once.
Use everywhere.
Supported environments include:
- Claude Code
- Codex
- Cursor
- Gemini CLI
- VS Code
Contorium acts as a shared runtime understanding layer rather than a tool-specific integration.
⸻
The Bigger Direction
This update reflects a broader shift in how we think about AI development.
Most AI tools focus on context windows.
Contorium focuses on continuity.
Not just what the AI sees right now.
But what the project has been doing over time.
The long-term vision remains unchanged:
AI conversations end.
Project understanding should not.
🌐 https://www.contorium.dev/
https://github.com/ContoriumLabs/contorium
Top comments (0)