Continue vs goose

Where goose wins

There is nobody left to acquire

goose used to be block/goose. It is now aaif-goose/goose, and the docs site says so in a banner: goose has moved to the Agentic AI Foundation, in a post dated 2026-04-07. The AAIF was formed by the Linux Foundation on 2025-12-09, anchored by MCP, goose and AGENTS.md. Block contributed the project instead of keeping it. That is the one structural answer to what happened to Continue, because an acquisition needs a seller, and a foundation-held project does not have one. It is not a guarantee that goose will be popular in three years. It is a guarantee that its future is not a line item in someone's exit.

The rules are written down

goose ships a GOVERNANCE.md, and it is the only project in this comparison that does. It names roles, contributors, maintainers and core maintainers, so "who can merge this" has a published answer. Set that against Continue's current state: 456 open pull requests, 218 filed since the final release of which 179 are still open, and three merged, the last two on 2026-07-21, all three of them teardown rather than development. Continue's licence was fine. What ran out was people with commit rights, and no document anywhere said how new ones would be appointed.

No meter to reprice

There is no goose subscription, no seat price and no goose-branded token gateway on its docs site. You bring your own key or point it at a local model, and its stated values put remote frontier models and local private models on the same footing. Continue arrived at the same place, but only because its billing was switched off on the way out. One of those is a design decision and the other is a side effect.

It is shipping

v1.46.0 landed on 2026-08-12, the repository was pushed on 2026-08-21, and it has 53,114 stars. Continue's last release was 2026-06-19 and its last commit 2026-07-21. For a tool you point at model APIs that change every few weeks, the release cadence is not a vanity metric; it is the thing that decides whether next quarter's model works.

Where Continue still wins

goose has not rebuilt the part of Continue most people actually touched. Continue was an extension inside VS Code and JetBrains, with a sidebar and inline completion; goose is a desktop app, a CLI and an API. There is no tab completion and no sidebar-chat parity. If your daily use of Continue was ghost text while typing, goose does not give that back, and no amount of foundation governance changes it. Be honest with yourself about which half of Continue you used before you spend a week on this.

And the incumbent genuinely comes out ahead in one specific case. An unmaintained client that never calls a server also never changes underneath you. If you run Continue on VS Code 2.1.0 with explicit model definitions in config.yaml and your own key or Ollama, you have a tool that is frozen by definition, and the final four days of commits made that the supported path on purpose: the login requirement was removed, Hub slugs in the default config were replaced with explicit model definitions, and the sign-in links were stripped. For an air-gapped or compliance-locked setup, "will never change" reads as a feature. It stops being one the day you need a security patch or a provider that did not exist in June.

Migrating Continue to goose

The path most teams take:

1. Inventory first. List what you had in Continue: model definitions in config.yaml, rules, prompts, MCP servers, and anything you pulled from the Hub. Model definitions and MCP servers have equivalents. Anything that came from hub.continue.dev has no export path at all, because that host has no DNS record and continue.dev/export returns 404; the 30-day window closed on 2026-07-15. Rewrite those from whatever you have locally.
2. Move the keys, not the config. goose is configured its own way; the swap is your provider keys and your local model endpoint, not a file you can port. Expect to re-express rules as goose instructions rather than translate them line by line.
3. Run both for one cycle. Keep Continue installed while you learn a terminal and desktop workflow. Nothing forces the deadline any more, which is the one advantage of a shutdown that already happened, and both tools bill through your own provider account, so the overlap costs you double tokens rather than double subscriptions.
4. What does not move. Inline completion. Your Continue chat history. Anything shared through the Hub. Say that out loud to the team before you start, so nobody discovers it in week two.