Repository navigation
Extension system: broad, VS Code-like extensibility for T3 Code #31141
ztomas-codes
started this conversation in
Ideas
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Problem
T3 Code covers a growing range of coding workflows, but users and teams will always need integrations and UI that the core app does not ship. Maintaining a fork for each customization makes updates harder, while adding every integration to core increases the maintenance burden for everyone.
I'd like a first-class Extension system, with the breadth and everyday experience of extensions in VS Code. The ambition is to make substantial features possible outside the core repository, including features maintainers have never anticipated.
What extensions should be able to contribute
For example, a GitLab extension could add a merge-request panel, pipeline status, palette commands, and agent tools. A deployment extension could add a project action, show deployment logs, and expose a tool the agent can use when asked to deploy. Neither should require patching the main app.
A VS Code-like extension experience
I would want the whole extension workflow, not only an internal plugin loader:
The reference is VS Code's extensibility model and developer experience. Compatibility with relevant existing VS Code extension APIs would be worth exploring, but this proposal does not assume arbitrary existing VSIX packages can already run unchanged.
Remote environments and clients
The extension model should account for T3 Code's architecture from the beginning:
Keep the host dependable
Broad extensibility needs boundaries that preserve the app's performance and reliability. I'd like an isolated extension host, lazy activation, scoped subscriptions, cancellation, resource diagnostics, and a way to disable a broken extension without losing access to the app. Extensions should clean up their views and subscriptions when deactivated.
Permissions should cover filesystem access, terminal execution, network access, conversation content and secret-provider access. Credentials should use supported secret APIs rather than being copied into extension settings. Extension operations should go through public services and respect environment grants, rather than modifying the database or replacing orchestration internals.
Direction to discuss
Would maintainers support an extension platform with this breadth as a long-term direction? Which contribution points would make a useful first release while leaving room for the wider system?
This is a product proposal requesting discussion and approval of direction. It is not an implementation PR. A first release could start with commands, panels, settings and environment-side tools, but I would want the API and packaging model designed to grow beyond that initial set.
Related: browser-extension support #13979 covers extensions inside the integrated browser. This proposal concerns extending T3 Code itself.
Prepared with GPT-6.1 Sol through the Codex harness in T3 Code.
All reactions