Best IDE Agent

An agentic coding assistant for VS Code powered by local LLMs via LM Studio (or any OpenAI-compatible endpoint). No cloud, no Copilot dependency.

This extension uses an editor-agnostic agent core under src/core/, which keeps the architecture modular and easier to evolve over time. See TODO.md for known gaps and near-term fixes.

Features

• Chat sidebar with streaming responses from your local model
• Agentic tool loop: file/dir search, regex + semantic code search, targeted search_replace, diagnostics/symbol lookup, git status/diff, file writes, and terminal commands
• Optional MCP tool bridge: connect stdio MCP servers and call external tools from agent runs
• Chat defaults to the right-hand auxiliary sidebar
• Editor-native file review: agent edits stage as pending diffs; open on demand with Review, then Accept/Reject/Accept All, plus one-click revert of the last accepted turn
• Approval flow: run_command and mcp_call_tool require explicit pre-execution approval
• Dedicated agent shell for run_command (POSIX and Windows PowerShell), with command + output shown in collapsible chat cards
• Ask/Agent/Composer chat modes: use Ask for read-only analysis, Agent for standard tool use, or Composer for structured plan-and-apply multi-file tasks
• Project guidance via .bestide/rules.md, plus task-scoped skills loaded with @skill:name from .bestide/skills/
• Graceful step-limit resume: continue or stop when bestIde.maxSteps is reached
• Optional Tab/ghost-text code completion via VS Code inline completions
• Model picker backed by GET /v1/models
• Works with any OpenAI-compatible server (LM Studio, Ollama, llama.cpp server, vLLM, ...)

Getting started

1. Set up LM Studio

1. Install LM Studio
2. Download a model with tool-use support. Recommended starting points:

• google/gemma-4-e4b - runs well on macbooks with 32 gb+ ram
• qwen2.5-coder-7b-instruct (good balance of speed and tool-calling quality)
• qwen3-8b or larger Qwen3 variants
• llama-3.1-8b-instruct

1. Open the Developer tab and start the local server (defaults to http://localhost:1234)
2. Load the model into the server

2. Install the extension

From a marketplace (after publish):

• Visual Studio Marketplace
• Open VSX

From source / VSIX:

git clone <repo-url>
cd best-ide
npm install
npm run install:local

That builds a VSIX and installs it into Cursor (preferred) or VS Code with --force. Then Developer: Reload Window from the Command Palette.

Manual equivalent:

1. Build a VSIX: npm run vsix → best-ide-agent-<version>.vsix in the project root
2. Install:
   • CLI: cursor --install-extension best-ide-agent-*.vsix --force (or code instead of cursor)
   • UI: Extensions view → ... → Install from VSIX...
3. Configure: open Settings, search for bestIde, and run Best IDE: Set API Key if your LM Studio server has authentication enabled (LM Studio > Developer > API tokens). The key is stored in VS Code Secret Storage.

To pick up code changes after editing the repo, re-run npm run install:local and reload the window. For day-to-day extension development (F5 + watch), see CONTRIBUTING.md. Publishing runbook: docs/PUBLISHING.md.

Settings

Use Best IDE: Set API Key and Best IDE: Clear API Key from the command palette to manage the default API token in VS Code Secret Storage.

Privacy and offline story

• By default, Best IDE is local-first: point bestIde.baseUrl (or your local backend profile) at a local OpenAI-compatible server such as LM Studio.
• Extension data stays on your machine: conversation threads and pending file-review state are saved in local VS Code extension state, and agent file edits happen in your workspace.
• Nothing is sent to Best IDE-managed cloud services because this project does not run one. Network traffic goes only to the model/MCP endpoints you configure.
• If you add non-local backends (for fallback, cost, or quality routing), requests for those operations are sent to those configured providers by design.

Rules and skills

• Add project-wide guardrails in .bestide/rules.md (automatically merged into the system prompt).
• Add reusable task instructions under .bestide/skills/.
• Reference a skill in chat with @skill:name (for example @skill:test-driven loads .bestide/skills/test-driven.md if it exists).
• Skills can also be folders containing SKILL.md (for example .bestide/skills/refactor/SKILL.md).

Contributing

If you want to work on the extension itself, see CONTRIBUTING.md for developer setup, test workflows, and architecture notes. Near-term fixes are tracked in TODO.md. Publishing is documented in docs/PUBLISHING.md.

Known limitations (MVP)

• Tool-calling quality varies a lot between local models; small models may emit malformed calls
• run_command uses a persistent shell session (POSIX SHELL / Windows PowerShell); on timeout or shell crash, the session is restarted for the next command
• Extension-host UI glue (panel.ts, webview runtime) has thinner automated coverage than src/core; prefer extracted helpers for new tests