Connect DEVUP AI to VS Code—No Extension Required

A developer detailed a method for connecting the DEVUP AI platform to Visual Studio Code using VS Code's native Custom Endpoint provider, introduced in stable version 1.122, without installing any extension. The setup relies on a bring-your-own-key configuration pointing chat and tool-calling requests at DEVUP AI's Chat Completions endpoint, with a dedicated API key stored as a secure reference in chatLanguageModels.json. The guide notes the approach does not automatically replace every AI capability inside the editor and that some features remain limited by platform-specific constraints.

Disclosure: Visual Studio Code is a third-party product. DEVUP AI is an independent platform and is not affiliated with or endorsed by its publisher. AI-assisted development does not have to begin with another extension, another proprietary workflow, or another foreign billing account. VS Code now includes a native Custom Endpoint provider that can connect its chat experience to a compatible API using your own key. For developers in Algeria, this creates a practical path: No additional VS Code extension is required for the setup described in this guide. This article walks through the configuration carefully, explains what each field controls, and separates what works today from features that still have platform-specific limitations. The Custom Endpoint provider became available in the stable version of VS Code with version 1.122 . It supports three API formats: This guide uses the Chat Completions format exposed by DEVUP AI. VS Code's Bring Your Own Key experience can power chat and tools without requiring a bundled AI subscription. However, this does not automatically replace every AI capability inside the editor. We will cover those boundaries later. You can verify the current behavior in the official VS Code language-model documentation https://code.visualstudio.com/docs/agent-customization/language-models and the VS Code 1.122 release notes https://code.visualstudio.com/updates/v1 122 . The native Language Models editor in VS Code. Source: Visual Studio Code documentation https://code.visualstudio.com/docs/agent-customization/language-models . By the end of this guide, you will have: DEVUP AI . The only public endpoint required for this integration is: https://api.devupai.com/v1/chat/completions No internal platform configuration is required. Before starting, prepare the following: You can obtain an API key from the DEVUP AI dashboard https://devupai.com/dashboard/api-keys and choose a model from the DEVUP AI model catalog https://devupai.com/models . For agent workflows, choose a model that explicitly supports tool calling . Confirm that you are running VS Code 1.122 or newer . On Windows and Linux: Help → Check for Updates On macOS: Code → Check for Updates Restart VS Code after the update. If Custom Endpoint does not appear later in the setup, checking the installed version should be your first troubleshooting step. Open the API Keys dashboard https://devupai.com/dashboard/api-keys and create a dedicated key for your VS Code environment. A dedicated key is preferable because it can be rotated or revoked independently. VS Code prompts for the key during the configuration flow. Keep the generated secret reference intact instead of replacing it with the raw value. Open the VS Code Command Palette: Ctrl + Shift + P Cmd + Shift + P Run: Chat: Manage Language Models You can also open the Chat view, select the current model, and choose Manage Language Models . Inside the Language Models editor: The provider picker used to add a Custom Endpoint. Source: Visual Studio Code documentation https://code.visualstudio.com/docs/agent-customization/language-models . VS Code will then open: chatLanguageModels.json The setup wizard creates the provider-level configuration and a secure reference to your API key. Do not replace that reference with your raw key. Inside the Custom Endpoint group's models array, add a model entry using the template below. The placeholders are intentionally not valid final values. Replace every value wrapped in <... with information from the selected model's DEVUP AI catalog page before saving. { "id": "<EXACT MODEL ID FROM DEVUP AI ", "name": "DEVUP AI Coding Model", "url": "https://api.devupai.com/v1/chat/completions", "toolCalling": true, "vision": false, "maxInputTokens": "<CONTEXT WINDOW MINUS MAX OUTPUT ", "maxOutputTokens": "<MODEL MAX OUTPUT TOKENS " } The two token fields must be JSON numbers, not strings, after replacing the placeholders. For example, the final form should follow this shape: { "maxInputTokens": 24000, "maxOutputTokens": 8000 } Those numbers are only a formatting example. They are not universal model limits. | Field | Purpose | |---|---| | id | The exact model identifier copied from the DEVUP AI catalog | | name | The label displayed inside the VS Code model picker | | url | The full DEVUP AI Chat Completions endpoint | | toolCalling | Enables workspace tools when the selected model supports them | | vision | Enables image input only when supported by the model | | maxInputTokens | Maximum input budget exposed to VS Code | | maxOutputTokens | Maximum generation budget supported by the model | VS Code treats the sum of the input and output limits as the total context window: maxInputTokens + maxOutputTokens ≤ model context window A safe configuration is: maxInputTokens = context window − maxOutputTokens Copy the model-specific limits from its catalog page. Do not assume that two models accept the same values. Set: "toolCalling": true only when the selected model supports tool calling. This property informs VS Code of an existing model capability. It cannot add tool support to a model that does not already provide it. Use: "vision": true only when the selected model supports image input. Otherwise, keep it set to false . Save chatLanguageModels.json , return to the Chat view, and open the model picker. Your configured model should appear under the DEVUP AI group. If it does not appear: Do not begin by asking the model to modify an entire repository. Start with a read-only request: Review the currently open file. Do not edit anything. Explain: 1. What the file does. 2. Its main dependencies. 3. One potential issue worth investigating. This verifies: If this test succeeds, continue to a controlled tool test. Open a small test repository and send: Inspect this project and identify its test command. Do not modify any files. Do not run the command yet. Show me the command you found and wait for my approval. A successful response should identify the relevant project file, locate the test command, and stop before execution. After reviewing the proposed command, approve it only if it is safe. This confirms that the model can participate in a controlled tool workflow without immediately making irreversible changes. Once the read-only and tool tests pass, try a constrained engineering task: Analyze the failing test in this repository. Before editing: 1. Explain the probable root cause. 2. List the files you intend to change. 3. Propose the smallest safe fix. 4. Wait for approval. After approval: 1. Apply the minimum required change. 2. Run only the relevant tests. 3. Report the exact validation result. 4. Summarize every modified file. This prompt creates explicit checkpoints between analysis, modification, and validation. The agent produces a candidate change, but you still own the decision to accept it. The VS Code Chat view beside an active workspace. Source: Visual Studio Code documentation https://code.visualstudio.com/docs/agents/overview . With a correctly configured and compatible model, the native Custom Endpoint path can provide: The exact result depends on the selected model's capabilities and the permissions granted inside VS Code. BYOK support is powerful, but its boundaries matter. The Custom Endpoint configuration described here powers chat and compatible tool workflows. It does not automatically replace the editor's inline code-completion system. Some semantic search and embedding-dependent capabilities can still require a separate account sign-in and are not powered by the selected BYOK chat model. The standard Chat view and the separate Agents window are not identical execution surfaces. BYOK support inside Agent Host sessions has a separate experimental setting and may change. Do not assume that every model configured for Chat behaves identically in every agent interface. In managed Business or Enterprise environments, an administrator can disable Bring Your Own Key through organization policy. If the Custom Endpoint option is missing on a managed device, contact the organization administrator before changing local configuration. | Symptom | Most likely cause | Correct action | |---|---|---| | Custom Endpoint is missing | VS Code is outdated or BYOK is restricted | Update to 1.122+ and check organization policy | | Model does not appear | Invalid JSON or unsaved configuration | Validate the file, save, and restart VS Code | | Authentication fails | Invalid, revoked, or incorrectly stored key | Update the API key through the model-provider settings | | Chat works but agent tools do not | Model lacks tool calling or the flag is disabled | Choose a tool-capable model and verify toolCalling | | Requests fail on long context | Incorrect context or output limits | Copy the correct limits from the model page | | Image prompts fail | vision does not match model capability | Set the field according to the selected model | | Endpoint returns a route error | An incomplete URL was configured | Use the full /v1/chat/completions endpoint | For integration-specific configuration, consult the DEVUP AI VS Code guide https://devupai.com/integrations/vscode . Before accepting an agent-generated change: The correct mental model is not: The agent wrote it, therefore it works. It is: The agent produced a candidate change and evidence. The developer still owns the final decision. The native Custom Endpoint integration is available today and provides the fastest path to using DEVUP AI inside VS Code without installing an additional extension. At the same time, we are actively building a dedicated DEVUP AI extension for VS Code . Unlike a generic endpoint adapter, this extension is being designed for full, end-to-end integration with DEVUP AI and a broader set of professional development capabilities directly inside the editor. The extension is still under active development. Its final feature set, compatibility matrix, beta availability, and release timeline will be announced after technical validation. The dedicated extension will complement the native Custom Endpoint workflow described in this guide—not make it obsolete. Developers who want immediate access can use the native integration today. Those who want the complete DEVUP AI-native editor experience can follow the platform for the upcoming extension announcement. VS Code's native Custom Endpoint support changes the integration equation. You can now connect DEVUP AI to the editor, select a compatible model, use streamed chat, validate tool calling, and keep billing in Algerian dinars—without installing an additional extension. The setup is small. The important part is configuring model capabilities accurately, validating them in stages, and keeping the developer in control of every consequential action. If you test the integration, share what you built and which VS Code workflow you want the dedicated DEVUP AI extension to improve first.