Installer Continue dans VS Code : extension, config, clés API, Ollama, édition inline
Installer Continue, clés API ou Ollama, Chat @ et un Edit inline relisible.
La première heure avec Continue open source dans VS Code bloque surtout sur l’extension Continue.continue, un config.yaml sans modèle, ou une clé API / Ollama local non connecté—not sur le prompt. Installez depuis le Marketplace, ajoutez un modèle dans le panneau Continue ou ~/.continue/config.yaml, puis essayez l’édition inline et le chat avec contexte @ codebase.
Requête : installer et utiliser Continue dans VS Code. Suit docs.continue.dev (VS Code, fournisseurs, Chat, Edit) et Continue.continue. Continue est un assistant OSS distinct, pas le produit réunion de ce site. Appel court : wbmeet, créer et rejoindre une salle.
Ce que Continue fait dans VS Code
Continue is an open-source (Apache 2.0) IDE assistant: sidebar Chat accepts @ files, folders, docs, and codebase context; Edit rewrites a selection inline with a reviewable diff; Autocomplete offers Tab completions when your model supports them. Extension id: Continue.continue. Models and API keys live in ~/.continue/config.yaml (or the Continue settings UI); Ollama and other local runners are supported. A Continue Hub account is optional—you can bring your own keys.
Installer l’extension
- 1
Install from VS Code Marketplace
Open VS Code,
Cmd+Shift+X/Ctrl+Shift+X, search Continue, confirm publisher and idContinue.continue, then Install. Continue also ships JetBrains plugins; this page covers VS Code only. - 2
Open sidebar and commands
Click the Continue icon in the Activity Bar, or use
Cmd+Shift+L/Ctrl+Shift+Lto focus Chat (shortcuts per docs.continue.dev/reference/keyboard-shortcuts and your build). Command Palette entries include Continue: Open Config, Edit, etc. - 3
First-run wizard (optional Hub)
First launch may offer model picks or Continue Hub sign-in. If you already have OpenAI, Anthropic, Gemini, or similar keys, choose bring-your-own-key or skip Hub and edit config directly. Confirm org policy before storing keys locally.
Configure API keys or local Ollama
- 1
Open config.yaml
In the Continue panel, open settings or run Continue: Open config.yaml to edit
~/.continue/config.yaml. The file listsmodelswithtitle,provider(openai, anthropic, ollama, …),model, andapiKeyor localapiBase. - 2
Cloud provider example
Anthropic: create a key at console.anthropic.com, set provider anthropic and a supported Claude model id. OpenAI and Gemini follow the same pattern from their consoles; billing and regions follow each vendor. Never paste live keys into repos or screenshots.
- 3
Local Ollama
Install and run Ollama locally,
ollama pulla model, then add a config entry with providerollama, model name, and default base URLhttp://localhost:11434. Continue docs describe Chat and Edit on local-only setups. - 4
Pick the active model
Use the model dropdown at the top of the Continue panel for Chat, Edit, and Autocomplete. After editing config, reload the window if the list is stale.
Chat and inline Edit
- 1
Chat with @ context
In Chat, use
@Files,@Code,@Codebase(exact names per current docs) to attach files or retrieved chunks. Hour one: read-only questions such as explaining exports in the current file. - 2
Inline Edit on a selection
Select a function or block, press
Cmd+I/Ctrl+I(or Continue: Edit), describe the change in plain language, review the diff, then accept. - 3
Project rules (optional)
Optional project files such as
.continuerules(see docs) steer Chat/Edit toward repo conventions. Keep secrets out of the workspace or in .gitignore.
# First hour · Continue · VS Code · 2026-09
# Extensions → Continue (Continue.continue)
# Continue: Open config.yaml → add model + apiKey or ollama
# Chat: @Files explain this module
# Select code → Cmd+I / Ctrl+I → small edit → accept diffFeature comparison
| Feature | Good for | First-hour note |
|---|---|---|
| Chat + @ | Read & plan | Read-only first |
| Edit (Cmd/Ctrl+I) | Local rewrites | Review diff before accept |
| Autocomplete | Tab completions | Model-dependent |
| Ollama | Offline / privacy | Pull model first |
| config.yaml | Multi-model | Keys stay out of Git |
Empty model list
Validate config.yaml syntax, provider ids, and keys; confirm Ollama listens on localhost:11434.
401 / rate limits
Rotate keys; check vendor quotas; try a smaller model for the first Edit.
@Codebase slow
Prefer @ single file or selection; narrow scope in large monorepos.
Official path: install Continue.continue → configure config.yaml (cloud keys or Ollama) → Chat with @ context → Cmd/Ctrl+I with reviewed diff. Trust docs.continue.dev and your installed version.
- Vs Cody / Roo Code
- Each VS Code AI extension configures differently; you can keep several. See our Cody install note and Roo Code for comparison.
- Is Continue Hub required?
- No. Bring-your-own keys or Ollama works; Hub adds hosted models and team features per official pricing.
- Not a meeting-product guide
- Only Continue’s first hour. Short whiteboard call: wbmeet and when not to open another heavy meeting app.
Questions à vérifier
Must I sign up for Continue?
Not when using your own API keys or Ollama. Hub is for hosted models and synced config per docs.continue.dev.
config.json or config.yaml?
Recent builds prefer YAML; if docs mention JSON, trust the file opened by Continue: Open config.
Edit vs asking Chat to change code?
Edit targets the selection with a tight diff; Chat suits multi-file discussion—still review each suggested change.
Must the first task touch production?
No. Start with read-only @ questions, then one reversible Cmd/Ctrl+I edit.