Bring your own model
An AI coding assistant that runs on your models
Chat, an agentic multi-file coder, codebase-aware context, and inline edits — powered by a local Ollamadaemon or any OpenAI-compatible endpoint. No vendor lock-in, no forced cloud.
Everything you'd expect, none of the lock-in
Streaming chat
Plan, Edit, and Agent modes, with named/persistent sessions.
Bring your own model
A local Ollama daemon or any OpenAI-compatible endpoint. API keys live in your OS keychain, never in plaintext settings.
Agentic multi-file coder
A universal JSON tool protocol that works even on smaller open-source models — approval gates, end-of-run batch diff review (rejected edits never touch disk), one-click revert, and batched multi-edit.
Codebase-aware context
@codebase semantic search, @file, @currentFile, @symbol, @open, @problems, @diff, @commits, @terminal, @url, @search, @repomap, and @os — bring in exactly the context you need.
Inline edit & inline chat
Select code, describe a change, review a red/green diff — or ask a quick question with Ctrl-I without opening the chat panel.
Ghost-text completions
Real Ollama fill-in-the-middle for FIM-trained code models (codellama, deepseek-coder, qwen2.5-coder, …), with a chat-based fallback for everything else.
MCP servers
Configure MCP servers once; their tools become available to the agent loop automatically, named mcp__<server>__<tool>.
Project rules & prompt files
A global .xpreiIDErules file, modular glob-scoped .xpreiIDE/rules/*.md files, and .xpreiIDE/prompts/*.md files that become your own /name slash commands.
One engine, every IDE
JetBrains and Eclipse share the same engine as the VS Code extension via a bundled local core process.
Visual Studio Code
Marketplace / .vsix — the full feature set.
JetBrains
IntelliJ IDEA, PyCharm, WebStorm, GoLand, and more — one plugin covers them all.
Eclipse
Builds cleanly against a real JDK/Maven/Tycho toolchain.
CLI
Headless xprei agent / xprei chat for CI or no-IDE workflows.