OpenRouter Copilot Model Manager
Browse 400+ OpenRouter models and put the ones you want into GitHub Copilot Chat.
Compare price, context and capabilities, pick a few, and apply them to the Copilot model picker in one click.
What it does
- Browse the whole OpenRouter catalog. Search by name, provider or ID, filter by Vision, Tools, Free and Reasoning, pick a provider, and sort by name, price, context or release date.
- Compare at a glance. Every model shows its capabilities, input and output price per million tokens, context window and max output, with small meters so you can compare while scanning. In an editor tab the list becomes a sortable table.
- Add to Copilot in one click. Select models, then press Apply to Copilot. They appear in the Copilot Chat model picker under OpenRouter, with no reload and no config file.
- Manage what is live. The Active tab lists every model in Copilot. Remove one, or set its default Thinking effort.
- Full agent mode in Copilot Chat. Tool calling, vision input, streamed reasoning, prompt caching, retries with backoff, and token usage and cost in the status bar.
- Works without a Copilot subscription. Web fetch, web search, semantic code search, chat titles and commit messages, and inline suggestions all run through OpenRouter. See Without a Copilot subscription.
- See which model answered. With a router such as
openrouter/auto, the status bar names the model that actually answered and its provider.
- Fits anywhere. The panel follows your VS Code theme (light, dark or high contrast) and works from a narrow sidebar up to a full editor tab.
Install
Search OpenRouter Copilot Model Manager in the VS Code Extensions view, or install a .vsix:
code --install-extension openrouter-copilot-model-manager-1.3.1.vsix
Requires VS Code 1.104+ and GitHub Copilot Chat.
Quick start
- Open the OpenRouter Copilot Model Manager icon in the activity bar.
- Click the key icon and paste your OpenRouter API key. Get one at openrouter.ai/settings/keys. It is stored in VS Code SecretStorage, never in a file.
- The catalog loads on its own; click the sync icon to refresh it.
- Find models: press
/ to search, use the filter chips, the provider menu and sort.
- Press Select on the models you want. They wait in the tray at the bottom of the panel.
- Press Apply to Copilot (or
Ctrl+Enter). Your models are now in the Copilot Chat model picker.
A model that is already in Copilot shows a green In Copilot button and an Active pill, so you never add it twice. Clicking In Copilot removes it again.
Active in Copilot
Reasoning models get a Thinking Effort submenu in the Copilot model picker, so you can set the reasoning depth per request. The Active tab sets the default for each model, and openrouterModelManager.defaultReasoningEffort sets a global fallback.
Keyboard
| Key |
Action |
/ |
Focus search |
Escape |
Clear search, or close the provider menu |
Ctrl+Enter / ⌘+Enter |
Apply drafts to Copilot |
← → |
Switch between Browse and Active |
↑ ↓ |
Move through the provider menu |
Without a Copilot subscription
Copilot Chat works with OpenRouter models without a paid Copilot plan. Several Copilot features still call GitHub's own services, though, and fail or disappear without a subscription. The extension replaces them with OpenRouter:
| Copilot feature |
Without a subscription |
What this extension adds |
Fetch a web page (#fetch) |
Fails with "Your subscription has ended" when you are signed in with a lapsed plan |
Fetch Web Page (OpenRouter) tool, #openrouterFetch. On by default, and Copilot's own fetch tool is left out of requests so models cannot pick the one that fails (replaceCopilotFetch). |
| Web search |
Copilot has none for OpenRouter models |
OpenRouter web search with sources under the answer. Run OpenRouter: Toggle Web Search. About $0.007 per search with Exa. |
Semantic code search (#codebase) |
Not offered to OpenRouter models |
Codebase Search (OpenRouter) tool, #openrouterCodebase. The first use asks before indexing the workspace; indexing a typical repository costs under a cent. .env and key files are never sent. |
| Chat titles, commit messages, rename suggestions, repairing failed edits |
Fail, or quietly do nothing |
Run OpenRouter: Choose Utility Model for Copilot (the extension also offers it once). It points chat.utilityModel and chat.utilitySmallModel at a small OpenRouter model. |
| Inline suggestions (ghost text) |
Not available |
Run OpenRouter: Toggle Inline Completions. Off by default, because every suggestion is a paid request. Codestral answers in about a second. |
Searching other GitHub repositories (github_repo) |
Fails |
Not covered. Add the GitHub MCP server with a personal access token. |
If Copilot still asks for a subscription, sign out of GitHub in VS Code (Accounts menu → your GitHub account → Sign Out). Copilot treats a signed-in account without a plan more strictly than no account at all: signed out, OpenRouter models are fully supported. Web fetch no longer needs this.
Some Copilot extras cannot be replaced from an extension and stay unavailable without a plan: thinking-section titles, the goal summary badge, "explain changes" for edits and dictation cleanup.
Commands
| Command |
Description |
OpenRouter: Browse Models |
Open the model browser in an editor tab |
OpenRouter: Set API Key |
Set or update your OpenRouter API key |
OpenRouter: Sync Models from API |
Fetch the latest model catalog |
OpenRouter: Choose Utility Model for Copilot |
Pick the model for chat titles, commit messages and edit repair |
OpenRouter: Toggle Web Search |
Let models search the web while answering |
OpenRouter: Toggle Inline Completions |
Inline code suggestions from an OpenRouter model |
OpenRouter: Rebuild Codebase Search Index |
Index the workspace again from scratch |
OpenRouter: Delete Codebase Search Index |
Delete this workspace's local search index |
Settings
All settings are under openrouterModelManager.*:
| Setting |
Default |
Description |
cache.ttlMinutes |
60 |
Minutes before the model cache is refreshed |
requestTimeoutSeconds |
60 |
Request timeout in seconds |
maxRetries |
3 |
Retries for rate limits and server errors |
defaultTemperature |
(model default) |
Temperature from 0.0 to 2.0 |
defaultMaxTokens |
(model default) |
Maximum output tokens |
defaultReasoningEffort |
(model default) |
Thinking effort when Copilot has not set one: none, minimal, low, medium, high, xhigh or max |
maxInputTokensOverride |
(auto) |
Caps the context size reported to Copilot, so compaction starts earlier |
apiEndpoint |
https://openrouter.ai/api/v1 |
Custom API endpoint or proxy |
enablePromptCaching |
true |
OpenRouter prompt caching; cuts input-token costs in agent mode |
enableStreamUsage |
true |
Request token usage while streaming; turn off if a provider answers invalid_json |
sanitizeBase64Content |
true |
Strip long base64 blobs from prompts so guardrails do not block them; images are not affected |
logLevel |
info |
Output-channel verbosity: debug, info, warn or error |
utilityModel |
(none) |
OpenRouter model for chat titles, commit messages and edit repair; set it with the command |
replaceCopilotFetch |
true |
Leave Copilot's fetch_webpage out of requests when the extension's fetch tool is available |
webSearch.enabled |
false |
Let models that call tools search the web |
webSearch.engine |
exa |
exa (about $0.007 per search), parallel (about $0.001, results can be stale), native or auto |
webSearch.maxResults |
3 |
Results per search |
webSearch.showSources |
true |
List the pages an answer used under it |
codebaseSearch.embeddingModel |
openai/text-embedding-3-small |
Embedding model for codebase search |
codebaseSearch.maxFiles |
3000 |
Most files to index per workspace |
inlineCompletions.enabled |
false |
Inline suggestions from OpenRouter |
inlineCompletions.model |
mistralai/codestral-2508 |
Model for inline suggestions |
inlineCompletions.debounceMs |
350 |
Typing pause before a suggestion is requested |
The API key can also come from the OPENROUTER_API_KEY environment variable when none is stored.
Supported capabilities
| Capability |
Status |
| Chat (text to text) |
✅ |
| Tool calling (agent mode) |
✅ |
| Web fetch, web search and codebase search without a Copilot subscription |
✅ |
| Utility model for titles and commit messages |
✅ |
| Inline suggestions (opt-in) |
✅ |
| Vision / image input |
✅ |
| Thinking / reasoning display |
✅ |
| Thinking effort picker |
✅ |
| Streaming |
✅ |
| Retry with backoff |
✅ |
| Usage statistics |
✅ |
Development
npm install
npm run watch # esbuild in watch mode, then press F5 for an Extension Development Host
npm test # unit tests
npm run build # production bundle
npm run package # build a .vsix
Credits
The Copilot model-picker Thinking Effort integration was adapted from @Irvingouj's fork (c2dc1de), with the model-id heuristic removed and without enabling proposed APIs, so this build stays installable from the Marketplace on stable VS Code.
License
MIT. See LICENSE.
| |
