diff --git a/tips/activate-last-pane-keybinding.mdx b/tips/activate-last-pane-keybinding.mdx new file mode 100644 index 0000000..62dcc23 --- /dev/null +++ b/tips/activate-last-pane-keybinding.mdx @@ -0,0 +1,48 @@ +--- +title: Jump to Last Pane with Keybinding +subtitle: Add a keyboard shortcut to focus the rightmost pane instantly +category: navigation +difficulty: beginner +tags: + - panes + - keybindings + - navigation + - workflow +prUrl: 'https://github.com/zed-industries/zed/pull/49853' +publishedAt: '2026-03-16' +updatedAt: '2026-03-16' +author: godruoyi +authorUrl: 'https://github.com/godruoyi' +--- + +The new `workspace::ActivateLastPane` action lets you bind a keyboard shortcut that always focuses the rightmost pane, providing a predictable navigation target. + +## How to Use + +Add this to your keymap.json: + +```json +{ + "bindings": { + "cmd-9": "workspace::ActivateLastPane" + } +} +``` + +Now pressing **`Cmd-9`** always jumps to the last (rightmost) pane in your current window. + +## Why This Matters + +**Predictable navigation**: Unlike numbered pane actions, this never creates new panes. + +**Stable target**: Always focuses the rightmost pane regardless of how many panes exist. + +**Complements existing shortcuts**: Works well with `Cmd-1` through `Cmd-8` for other panes. + +**No side effects**: Pure focus action - never splits or modifies your layout. + +## Comparison with ActivatePane + +`workspace::ActivatePane` with an index can create/split panes when the index doesn't exist. `ActivateLastPane` is focus-only and never modifies your pane layout. + +This is perfect for users who want a reliable "jump to rightmost pane" shortcut similar to terminal or browser tab navigation patterns. diff --git a/tips/ctrl-enter-newline-search.mdx b/tips/ctrl-enter-newline-search.mdx new file mode 100644 index 0000000..95985c9 --- /dev/null +++ b/tips/ctrl-enter-newline-search.mdx @@ -0,0 +1,47 @@ +--- +title: Insert Newlines in Search Bars with Ctrl-Enter +subtitle: Add line breaks in search queries using keyboard shortcut on macOS +category: shortcuts +difficulty: beginner +tags: + - search + - keybindings + - multiline + - macos +prUrl: 'https://github.com/zed-industries/zed/pull/50420' +publishedAt: '2026-03-16' +updatedAt: '2026-03-16' +author: godruoyi +authorUrl: 'https://github.com/godruoyi' +--- + +On macOS, you can now use **`Ctrl-Enter`** to insert newlines in search bars, making it easier to search for multi-line patterns. + +## How to Use + +1. Open buffer search (**`Cmd-F`**) or project search (**`Cmd-Shift-F`**) +2. Type your search query +3. Press **`Ctrl-Enter`** to insert a newline +4. Continue typing on the next line +5. **`Enter`** still executes the search as before + +This keybinding is macOS-specific and matches common text editing patterns. + +## Why This Matters + +**Multi-line searches**: Find patterns that span multiple lines. + +**Regex patterns**: Format complex regular expressions for readability. + +**Code blocks**: Search for code snippets with proper line breaks. + +**Natural workflow**: **`Ctrl-Enter`** for newline, **`Enter`** for action. + +## Example Use Cases + +- Searching for multi-line function signatures +- Finding code blocks with specific structure +- Locating formatted text patterns +- Building readable regex patterns with line breaks for clarity + +Previously, there was no easy way to insert newlines in search bars - pressing **`Enter`** would execute the search. This enhancement brings better multi-line search support to Zed. diff --git a/tips/customize-vim-yank-highlight.mdx b/tips/customize-vim-yank-highlight.mdx new file mode 100644 index 0000000..9cbe02f --- /dev/null +++ b/tips/customize-vim-yank-highlight.mdx @@ -0,0 +1,56 @@ +--- +title: Customize Vim Yank Highlight Color +subtitle: Set a custom background color for vim yank highlighting in your theme +category: vim +difficulty: beginner +tags: + - vim + - theme + - customization + - visual +prUrl: 'https://github.com/zed-industries/zed/pull/49517' +mediaType: image +mediaUrl: 'https://cat.zed.tips/2026-03-16/customize-vim-yank-highlight-146fe90f.png' +publishedAt: '2026-03-16' +updatedAt: '2026-03-16' +author: godruoyi +authorUrl: 'https://github.com/godruoyi' +--- + +Vim mode now supports customizing the background color that appears when you yank (copy) text, allowing you to make it more visible or match your theme preferences. + +## How to Configure + +Add this to your settings.json: + +```json +{ + "theme_overrides": { + "Your Theme Name": { + "vim.yank.background": "#FE7F1F" + } + } +} +``` + +Replace `"Your Theme Name"` with your active theme (e.g., "Gruvbox Material", "One Dark", etc.) and choose your preferred color. + +## Why This Matters + +**Better visibility**: Make yank highlighting stand out more than the default document highlight color. + +**Theme consistency**: Match the yank highlight to your color scheme. + +**Accessibility**: Use higher contrast colors if needed for better visibility. + +**Personal preference**: Choose colors that work best for your eyes and workflow. + +Previously, vim yank highlighting was hardcoded to use `editor.document_highlight.read_background`, making it impossible to distinguish from other highlights. Now you have full control over this visual feedback. + +## Example Colors + +Try these popular choices: +- Orange: `#FE7F1F` (high visibility) +- Yellow: `#FFEB3B` (classic highlight) +- Blue: `#2196F3` (cool tone) +- Green: `#4CAF50` (subtle) diff --git a/tips/diagnostic-count-badges.mdx b/tips/diagnostic-count-badges.mdx new file mode 100644 index 0000000..5664e0d --- /dev/null +++ b/tips/diagnostic-count-badges.mdx @@ -0,0 +1,51 @@ +--- +title: See Error and Warning Counts in Project Panel +subtitle: Display diagnostic badges next to files showing error and warning counts +category: productivity +difficulty: beginner +tags: + - diagnostics + - project-panel + - errors + - ui +prUrl: 'https://github.com/zed-industries/zed/pull/49802' +mediaType: image +mediaUrl: 'https://cat.zed.tips/2026-03-16/diagnostic-count-badges-c9bb5860.png' +publishedAt: '2026-03-16' +updatedAt: '2026-03-16' +author: godruoyi +authorUrl: 'https://github.com/godruoyi' +--- + +The project panel can now show colored badges displaying the number of errors and warnings for each file, making it easy to spot which files need attention at a glance. + +## How to Enable + +1. Open Settings (**`Cmd-,`** on macOS) +2. Search for **diagnostic_badges** +3. Enable the option +4. Error counts appear in red, warnings in yellow next to file names + +Diagnostic counts automatically bubble up to parent directories, showing the total count for all files within. + +## Why This Matters + +**Visual scanning**: Quickly identify problematic files without opening them. + +**Prioritize work**: See which files have the most issues at a glance. + +**Track progress**: Watch error counts decrease as you fix issues. + +**Directory-level view**: Parent folders show aggregate counts for all contained files. + +When enabled, diagnostic severity takes priority over git status for file text color, while git status remains visible through file icon decorations. + +## Configuration + +The setting is disabled by default. Add to your settings.json: + +```json +{ + "diagnostic_badges": true +} +``` diff --git a/tips/git-panel-diff-stats.mdx b/tips/git-panel-diff-stats.mdx new file mode 100644 index 0000000..e4302fe --- /dev/null +++ b/tips/git-panel-diff-stats.mdx @@ -0,0 +1,53 @@ +--- +title: See Addition and Deletion Counts in Git Panel +subtitle: View file-level diff statistics showing lines added and removed +category: git +difficulty: beginner +tags: + - git + - diff + - statistics + - ui +prUrl: 'https://github.com/zed-industries/zed/pull/49519' +mediaType: image +mediaUrl: 'https://cat.zed.tips/2026-03-16/git-panel-diff-stats-d6ecc2df.png' +publishedAt: '2026-03-16' +updatedAt: '2026-03-16' +author: godruoyi +authorUrl: 'https://github.com/godruoyi' +--- + +The git panel now displays diff statistics for each modified file, showing you exactly how many lines were added and removed without opening the file. + +## How to Enable + +1. Open Settings (**`Cmd-,`** on macOS) +2. Search for **git_panel** +3. Find the **diff_stats** option and enable it +4. Addition and deletion counts appear next to each file in the git panel + +The counts are shown as colored numbers: green for additions, red for deletions. + +## Why This Matters + +**Quick assessment**: Gauge the scope of changes at a glance. + +**Review planning**: Prioritize which files to review based on change size. + +**Spot outliers**: Identify unexpectedly large changes that might need extra attention. + +**Context awareness**: Understand the impact of commits before viewing detailed diffs. + +## Configuration + +Add to your settings.json: + +```json +{ + "git_panel": { + "diff_stats": true + } +} +``` + +The setting is disabled by default to keep the git panel minimal, but enabling it provides valuable context for code review and commit planning. diff --git a/tips/jump-to-file-from-agent-diff.mdx b/tips/jump-to-file-from-agent-diff.mdx new file mode 100644 index 0000000..3c4708d --- /dev/null +++ b/tips/jump-to-file-from-agent-diff.mdx @@ -0,0 +1,48 @@ +--- +title: Jump to Files from Agent Diffs +subtitle: Open files directly from agent-shown diffs with Option-Enter +category: ai +difficulty: beginner +tags: + - agent + - navigation + - diff + - workflow +prUrl: 'https://github.com/zed-industries/zed/pull/50292' +mediaType: video +mediaUrl: 'https://cat.zed.tips/2026-03-16/jump-to-file-from-agent-diff-1a16202b.mov' +publishedAt: '2026-03-16' +updatedAt: '2026-03-16' +author: godruoyi +authorUrl: 'https://github.com/godruoyi' +--- + +When reviewing diffs in the agent conversation, you can now jump directly to the actual file by placing your cursor in the diff and pressing **`Option-Enter`**. + +## How to Use + +1. Review agent changes in the Agent Panel +2. Click anywhere in a diff block to place your cursor +3. Press **`Option-Enter`** to open the file at that location +4. Use **`Shift-Option-Enter`** to open in a split pane + +This uses the standard `editor: Open excerpts` and `editor: Open excerpts split` actions. + +## Why This Matters + +**Quick navigation**: Jump from conversation to code instantly. + +**Context switching**: Move seamlessly between review and editing. + +**Efficient workflow**: No need to manually search for files mentioned in diffs. + +**Split view support**: Open side-by-side for simultaneous comparison. + +## Example Use Cases + +- Jump to a file to review agent changes in full context +- Navigate to related code after understanding agent's modifications +- Open files in split view to compare before/after states +- Quickly investigate why the agent made certain changes + +This makes the agent conversation feel more integrated with your editing workflow, reducing friction when moving between discussion and implementation. diff --git a/tips/manage-built-in-agents.mdx b/tips/manage-built-in-agents.mdx new file mode 100644 index 0000000..9a5ad91 --- /dev/null +++ b/tips/manage-built-in-agents.mdx @@ -0,0 +1,53 @@ +--- +title: Remove and Manage Built-in Agents +subtitle: All built-in agents are now removable via the ACP Registry +category: ai +difficulty: beginner +tags: + - agent + - acp + - registry + - customization +prUrl: 'https://github.com/zed-industries/zed/pull/50094' +publishedAt: '2026-03-16' +updatedAt: '2026-03-16' +author: godruoyi +authorUrl: 'https://github.com/godruoyi' +--- + +All of Zed's built-in agents (Claude, Codex, Gemini, etc.) have been migrated to the ACP Registry, which means you can now remove agents you don't use and manage them like any other extension. + +## What Changed + +Previously, built-in agents were hardcoded and always available. Now they're: +- Listed in the ACP Registry alongside community agents +- Removable if you don't need them +- Reinstallable from the registry at any time +- Managed through the same interface as third-party agents + +## How to Manage Agents + +1. Open the Agent Panel +2. Click on agent settings or the registry icon +3. View all installed agents (including built-in ones) +4. Remove unwanted agents with the uninstall button +5. Reinstall from the registry if needed later + +## Why This Matters + +**Cleaner workspace**: Remove agents you never use from the interface. + +**Reduced clutter**: Agent picker only shows relevant options. + +**User control**: Treat built-in agents like any other extension. + +**Consistent management**: One interface for all agents regardless of source. + +## Example Use Cases + +- Remove Gemini if you only use Claude +- Uninstall Codex if you prefer other models +- Simplify the agent picker to just your preferred agents +- Reinstall agents when your workflow changes + +This change gives you complete control over your agent setup, letting you tailor Zed's AI features to your specific needs and preferences. diff --git a/tips/parallel-subagents.mdx b/tips/parallel-subagents.mdx new file mode 100644 index 0000000..af6c58a --- /dev/null +++ b/tips/parallel-subagents.mdx @@ -0,0 +1,61 @@ +--- +title: Parallel Subagents for Complex Tasks +subtitle: >- + Agent can now spawn parallel subagents for better context management and + concurrent work +category: ai +difficulty: intermediate +tags: + - agent + - subagents + - parallel + - performance +prUrl: 'https://github.com/zed-industries/zed/pull/50493' +publishedAt: '2026-03-16' +updatedAt: '2026-03-16' +author: godruoyi +authorUrl: 'https://github.com/godruoyi' +--- + +Zed's AI agent can now spawn multiple subagents that work in parallel, allowing it to tackle complex tasks more efficiently while managing context better. + +## What Are Subagents? + +Subagents are separate AI agent instances that the main agent can create to: +- Work on independent subtasks simultaneously +- Keep focused context for specific problems +- Prevent context window overflow on large tasks +- Parallelize research and implementation work + +The main agent coordinates the subagents and combines their results. + +## How It Works + +When you give the agent a complex task: + +1. The agent analyzes the task and identifies independent subtasks +2. It spawns subagents using the new `spawn_agent` tool +3. Subagents work in parallel on their assigned subtasks +4. The main agent collects and integrates the results +5. You get faster completion for multi-part tasks + +This happens automatically - you don't need to request subagents explicitly. + +## Why This Matters + +**Faster completion**: Parallel work reduces total time for complex tasks. + +**Better context**: Each subagent maintains focused context for its subtask. + +**Scalability**: Handle larger tasks without overwhelming a single agent's context. + +**Smarter coordination**: Main agent orchestrates work like a senior developer delegating to team members. + +## Example Use Cases + +- Large refactoring across multiple files +- Researching multiple code patterns simultaneously +- Implementing independent features in parallel +- Analyzing different aspects of a codebase concurrently + +The feature is now available to all users after being tested behind a feature flag. diff --git a/tips/repl-clear-outputs.mdx b/tips/repl-clear-outputs.mdx new file mode 100644 index 0000000..57576b1 --- /dev/null +++ b/tips/repl-clear-outputs.mdx @@ -0,0 +1,66 @@ +--- +title: Clear REPL Outputs with Commands +subtitle: Remove cell outputs to declutter notebooks or start fresh +category: productivity +difficulty: beginner +tags: + - repl + - notebook + - commands + - workflow +prUrl: 'https://github.com/zed-industries/zed/pull/49631' +publishedAt: '2026-03-16' +updatedAt: '2026-03-16' +author: godruoyi +authorUrl: 'https://github.com/godruoyi' +--- + +The REPL now includes commands to clear output from notebook cells, helping you declutter your workspace or prepare notebooks for sharing. + +## Available Commands + +Two new commands are available via the command palette: + +**`repl: Clear Current Output`** +- Clears output from the currently selected cell +- Useful for removing stale or unwanted results +- Keeps the code intact + +**`repl: Clear Outputs`** +- Clears outputs from all cells in the notebook +- Perfect for cleaning up before committing +- Resets the notebook to a clean state + +## How to Use + +1. Open a notebook file (`.ipynb`) in Zed +2. Open the command palette (**`Cmd-Shift-P`**) +3. Search for "clear output" or "clear outputs" +4. Select the desired command + +You can also bind these commands to custom keybindings in your keymap.json. + +## Why This Matters + +**Clean commits**: Remove outputs before committing notebooks to version control. + +**Declutter workspace**: Remove large or distracting outputs while working. + +**Privacy**: Clear sensitive data from outputs before sharing. + +**Start fresh**: Reset all outputs when re-running entire notebooks. + +## Example Keybindings + +Add to your keymap.json: + +```json +{ + "bindings": { + "cmd-k cmd-o": "repl::ClearCurrentOutput", + "cmd-k shift-o": "repl::ClearOutputs" + } +} +``` + +This feature addresses a common notebook workflow need that was previously missing from Zed's REPL interface. diff --git a/tips/repl-remote-kernels.mdx b/tips/repl-remote-kernels.mdx new file mode 100644 index 0000000..fa76769 --- /dev/null +++ b/tips/repl-remote-kernels.mdx @@ -0,0 +1,56 @@ +--- +title: Use REPL with WSL and SSH Remote Kernels +subtitle: Run Jupyter kernels on remote machines or WSL environments +category: productivity +difficulty: intermediate +tags: + - repl + - remote + - wsl + - jupyter +prUrl: 'https://github.com/zed-industries/zed/pull/47891' +publishedAt: '2026-03-16' +updatedAt: '2026-03-16' +author: godruoyi +authorUrl: 'https://github.com/godruoyi' +--- + +The REPL now supports running Jupyter kernels on WSL (Windows Subsystem for Linux) and SSH remote machines, letting you execute code in remote environments directly from Zed. + +## What Are Remote Kernels? + +Remote kernels allow you to run notebook code on a different machine or environment than where Zed is running. This is useful for: +- Using powerful remote servers for computations +- Running code in Linux environments from Windows via WSL +- Accessing GPU resources on remote machines +- Working with specific environments or dependencies + +## How to Use + +1. Open a notebook file (`.ipynb`) in Zed +2. Click the kernel selector in the REPL interface +3. Choose from available kernel options: + - **Local kernels**: Run on your machine + - **WSL kernels**: Run in Windows Subsystem for Linux + - **SSH kernels**: Run on remote SSH servers +4. Select your desired kernel and start running code + +The kernel spawning and lifecycle management is handled automatically through the remote protocol. + +## Why This Matters + +**Remote computing**: Leverage powerful servers without leaving Zed. + +**Environment isolation**: Run code in specific environments (WSL, Docker, etc.). + +**Cross-platform**: Windows users can use Linux environments seamlessly. + +**Resource access**: Connect to machines with GPUs, large datasets, or specific software. + +## Requirements + +- **WSL**: Windows Subsystem for Linux must be installed and configured +- **SSH**: SSH access to remote machines with Jupyter kernels installed +- **Kernels**: Target environment must have required Jupyter kernels available + +This feature integrates with Zed's existing remote development capabilities, bringing the full power of remote computing to notebook workflows. diff --git a/tips/review-diff-with-agent.mdx b/tips/review-diff-with-agent.mdx new file mode 100644 index 0000000..c599f54 --- /dev/null +++ b/tips/review-diff-with-agent.mdx @@ -0,0 +1,42 @@ +--- +title: Review Branch Diffs with Agent +subtitle: Send entire git diffs to your AI agent for automated code review +category: git +difficulty: beginner +tags: + - git + - agent + - code-review + - diff +prUrl: 'https://github.com/zed-industries/zed/pull/49513' +mediaType: image +mediaUrl: 'https://cat.zed.tips/2026-03-16/review-diff-with-agent-c9312b11.png' +publishedAt: '2026-03-16' +updatedAt: '2026-03-16' +author: godruoyi +authorUrl: 'https://github.com/godruoyi' +--- + +When reviewing your branch changes before committing or creating a PR, you can now send the entire diff to your AI agent for an automated code review with a single click. + +## How to Use + +1. Open the command palette and run **`git: branch diff`** +2. Review your changes in the split diff view +3. Click the **Review Branch** button at the top +4. The entire diff is automatically sent to your last-used agent +5. Review the agent's feedback in the Agent Panel + +The agent receives a pre-written prompt asking it to review your changes for potential issues, improvements, and best practices. + +## Why This Matters + +**Quick feedback**: Get instant code review suggestions without waiting for teammates. + +**Catch issues early**: Identify potential bugs or improvements before committing. + +**Learn best practices**: The agent explains why certain changes might be problematic. + +**Seamless workflow**: No need to copy/paste diffs or switch contexts. + +This feature is perfect for solo developers, quick sanity checks, or getting a second opinion before requesting human review. diff --git a/tips/smart-multi-cursor-paste.mdx b/tips/smart-multi-cursor-paste.mdx new file mode 100644 index 0000000..d83b201 --- /dev/null +++ b/tips/smart-multi-cursor-paste.mdx @@ -0,0 +1,58 @@ +--- +title: Smart Multi-Cursor Paste Distribution +subtitle: Paste multiple lines to multiple cursors with automatic line-to-cursor mapping +category: productivity +difficulty: intermediate +tags: + - multi-cursor + - clipboard + - editing + - workflow +prUrl: 'https://github.com/zed-industries/zed/pull/48676' +publishedAt: '2026-03-16' +updatedAt: '2026-03-16' +author: godruoyi +authorUrl: 'https://github.com/godruoyi' +--- + +When you have multiple cursors and paste content with the same number of lines, Zed now intelligently distributes each line to its corresponding cursor. + +## How It Works + +1. Create multiple cursors (e.g., 3 cursors) +2. Copy content with the same number of lines (e.g., 3 lines) +3. Paste with **`Cmd-V`** +4. Each line goes to its matching cursor position + +**Example:** + +If you copy: +``` +apple +banana +cherry +``` + +And paste with 3 cursors, you get: +- Line 1 ("apple") → Cursor 1 +- Line 2 ("banana") → Cursor 2 +- Line 3 ("cherry") → Cursor 3 + +## Why This Matters + +**Batch transformations**: Apply different values to multiple locations simultaneously. + +**Data entry**: Quickly distribute lists across multiple positions. + +**Refactoring**: Replace multiple items with different values in one paste. + +**Natural workflow**: Works intuitively when cursor and line counts match. + +If the line count doesn't match the cursor count, Zed falls back to the standard paste behavior (same content at each cursor). + +## Example Use Cases + +- Distributing different variable names to multiple function parameters +- Pasting multiple import statements to different locations +- Filling in template placeholders with different values +- Renaming multiple similar elements with distinct names diff --git a/tips/undo-reject-all-agent-changes.mdx b/tips/undo-reject-all-agent-changes.mdx new file mode 100644 index 0000000..ff08e3d --- /dev/null +++ b/tips/undo-reject-all-agent-changes.mdx @@ -0,0 +1,41 @@ +--- +title: Undo "Reject All" in Agent Panel +subtitle: Recover agent edits if you accidentally reject all changes +category: ai +difficulty: beginner +tags: + - agent + - undo + - edits + - workflow +prUrl: 'https://github.com/zed-industries/zed/pull/48462' +mediaType: video +mediaUrl: 'https://cat.zed.tips/2026-03-16/undo-reject-all-agent-changes-e3d942c1.mov' +publishedAt: '2026-03-16' +updatedAt: '2026-03-16' +author: godruoyi +authorUrl: 'https://github.com/godruoyi' +--- + +Accidentally clicked **Reject All** on agent changes? You can now undo this action and recover all the edits the agent made. + +## How to Use + +1. After clicking **Reject All** on agent changes +2. A toast notification appears with an **Undo** button +3. Click **Undo** to restore all rejected changes +4. All agent edits are re-applied to your files + +You can also use the standard undo keybinding (**`Cmd-Z`**) immediately after rejecting. + +## Why This Matters + +**Safety net**: Prevents accidental data loss from misclicks. + +**Second chances**: Review changes again if you change your mind. + +**Confidence**: Reject changes freely knowing you can undo. + +**Faster workflow**: No need to re-run the agent to get changes back. + +This feature gives you peace of mind when reviewing large batches of agent-generated changes, knowing you can always recover if you make a mistake. diff --git a/tips/vercel-ai-gateway-provider.mdx b/tips/vercel-ai-gateway-provider.mdx new file mode 100644 index 0000000..51144ca --- /dev/null +++ b/tips/vercel-ai-gateway-provider.mdx @@ -0,0 +1,71 @@ +--- +title: Use Vercel AI Gateway as LLM Provider +subtitle: Connect to multiple AI models through Vercel's unified AI Gateway +category: ai +difficulty: intermediate +tags: + - ai + - providers + - vercel + - configuration +prUrl: 'https://github.com/zed-industries/zed/pull/50207' +publishedAt: '2026-03-16' +updatedAt: '2026-03-16' +author: godruoyi +authorUrl: 'https://github.com/godruoyi' +--- + +Zed now supports Vercel AI Gateway as a language model provider, giving you access to multiple AI models through Vercel's unified gateway interface. + +## What is Vercel AI Gateway? + +Vercel AI Gateway is a service that provides a single API endpoint for multiple AI model providers. Instead of configuring each provider separately, you connect to the gateway which handles routing to different models. + +## How to Configure + +Add to your settings.json: + +```json +{ + "language_models": { + "vercel_ai_gateway": { + "available_models": [ + { + "name": "gpt-4", + "provider": "openai" + } + ] + } + } +} +``` + +Then authenticate with your Vercel AI Gateway credentials when prompted. + +## Why Use AI Gateway? + +**Unified access**: Connect to multiple model providers through one endpoint. + +**Cost tracking**: Centralized usage monitoring and billing. + +**Rate limiting**: Built-in rate limit management across providers. + +**Model switching**: Easily switch between different AI models. + +**Enterprise features**: Advanced logging, caching, and fallback handling. + +## Supported Capabilities + +The provider automatically detects model capabilities from tags: +- Tool use (function calling) +- Vision (image understanding) +- Streaming responses + +Models are configured with their supported parameters based on the underlying provider. + +## Use Cases + +- Teams using Vercel's infrastructure for unified AI model access +- Projects requiring multi-model support with centralized management +- Organizations needing detailed usage analytics and cost attribution +- Developers who want simplified configuration for multiple AI providers