Feature request: AI-powered inline code completion in Code Editor (like Copilot)

[LJIAH]

Summary

Add AI-powered real-time inline code completion (ghost text / Copilot-style suggestions) to Warp's Code Editor. When typing code, the AI would predict and show gray ghost text of the likely next code, accepting with Tab or other shortcut.

Problem

Warp's Code Editor currently supports LSP for hover info, go-to-definition, diagnostics, and format-on-save, but lacks real-time code autocomplete — both LSP-based completions and AI-powered inline completions. Developers writing or editing code in Warp's editor need to switch to another IDE (VS Code, Cursor, etc.) to get the fluid autocomplete experience. This breaks the flow that Warp's "Agentic Development Environment" aims to create.

A related issue (#7346) was closed when LSP support shipped, but the "autocomplete" part was never implemented: the current LSP implementation covers hover, go-to-definition, find references, inline diagnostics, and format-on-save — but not textDocument/completion.

Desired workflow

1. Open a code file in Warp's Code Editor
2. Start typing code (e.g., console.)
3. AI predicts the next tokens and shows ghost text inline (grayed out) after the cursor
4. Press Tab (or another configurable key) to accept the suggestion
5. The suggestion may be multi-token / multi-line, predicting variable names, function calls, imports, etc.

Artifacts

None attached

Warp version

Unknown

Operating system

Windows (bash)

[oz-for-oss]

@LJIAH

Here's what I found while triaging this issue:

• This appears to be a feature gap in Warp's Code Editor rather than a bug: current editor/LSP support does not include autocomplete or AI ghost-text suggestions.
• Add LSP-backed autocomplete with auto-import to the code editor #9548 is related for LSP-backed Code Editor autocomplete, but this request is broader because it asks specifically for AI-powered inline suggestions.

Maintainer details

I concluded that feature request for AI-powered inline ghost-text code completion in Warp's Code Editor; current code and related issues support this as an unimplemented editor/AI suggestions capability rather than a bug..

Triage summary

This is an enhancement request for Copilot-style inline code completion in Warp's Code Editor. The request is actionable for product/design scoping: it asks for real-time AI ghost text while editing files, with keyboard acceptance and potentially multi-line suggestions.

Evidence

• The report gives a concrete desired flow: open a file in the Code Editor, type code such as console., show gray inline predicted text, and accept it with Tab or another shortcut.
• Current LSP code exposes document lifecycle, diagnostics, go-to-definition, formatting, hover, and find-references paths, but no completion request method in crates/lsp/src/model.rs:538, crates/lsp/src/model.rs:694, crates/lsp/src/model.rs:703, and crates/lsp/src/model.rs:721.
• app/src/code/local_code_editor.rs (990-1070) uses LSP primarily for format-on-save in the inspected save path; it does not provide an editor-completion request/response path.
• app/src/code/editor/view.rs (277-439) initializes the editor view, search, navigation, comments, and Vim behavior, but the inspected Code Editor surface does not include completion popup or ghost-text plumbing.

Duplicate check

I enumerated all open non-PR issues in warpdotdev/warp and found #9548 as a related open request for LSP-backed Code Editor autocomplete with auto-imports. I am not marking this issue as a duplicate because #9548 is focused on LSP popup autocomplete, while this issue specifically asks for AI-powered inline ghost-text suggestions, and the dedupe workflow requires at least two likely existing duplicates. The referenced #7346 is closed and outside the recent-closed window used for duplicate detection.

Likely implementation areas

• LSP completion support, if this feature includes standard textDocument/completion.
• Code Editor rendering/input support for inline ghost text and accepting/rejecting suggestions.
• AI suggestions/product policy for provider/model choice, latency, privacy, and how AI suggestions coexist with LSP completions.
• UI mocks are likely needed because this introduces visible inline editor affordances and keybinding behavior.

This is my automated analysis and may be incorrect. A maintainer will verify the details.

Powered by Oz