Dev Tool Configs

Privacy-focused configurations for AI coding assistants and editors. The default posture in this repo is:

• keep core AI features working
• disable telemetry and optional sharing where supported
• avoid recommendations that obviously break normal tool usage
• document vendor dashboard/account privacy controls alongside local config files

This is a best-effort reference repo, not a privacy guarantee.

What This Repo Covers

Tool	What's Here	Link
Claude Code	Balanced settings, privacy tiers, platform privacy controls	claude-code/
VS Code + Copilot	Editor telemetry settings, Copilot platform settings, agent-safety defaults	vscode/
Cursor	VS Code-layer settings, privacy mode guidance, remote-agent caveats	cursor/
OpenCode	Provider routing, OpenRouter privacy guidance, model examples	opencode/
Codex CLI	Sandbox/approval defaults, telemetry controls, local-history tradeoffs	codex-cli/
T3 Code	PostHog telemetry opt-out, underlying agent privacy guidance	t3-code/
Zed	Telemetry controls, agent permissions, provider privacy notes	zed/

Common Tool Folder Pattern

Each tool folder follows the same visible pattern where possible:

• README.md explains what the folder contains and how to install from the repo root
• one primary config file at the tool root is the default repo config
• one annotated companion beside it explains the defaults and recommendations
• examples/ exists only when a tool has real alternate profiles
• skills/ and skills-community/ exist only when a tool has a maintained skills pack

Tool	Default config	Annotated doc	Optional extras
Claude Code	settings.json	settings-annotated.md	skills/, skills-community/, configuration-guide.md
Codex CLI	config.toml	config-annotated.md	skills/, skills-community/
OpenCode	opencode.json	opencode-annotated.md	examples/, skills/, skills-community/
VS Code	settings.json	settings-annotated.md	none
Cursor	settings.json	settings-annotated.md	none
Zed	settings.json	settings-annotated.md	none

Skills

The repo keeps skills in visible per-tool folders:

• claude-code/skills/ for Claude-specific skills
• codex-cli/skills/ as the maintained source library for the shared Codex/OpenCode-compatible set
• opencode/skills/ as the native OpenCode-ready copy of the shared set plus any true OpenCode-only variants
• docs/skills-architecture.md as the canonical placement guide

Contributor favorites live in:

• codex-cli/skills-community/
• opencode/skills-community/
• claude-code/skills-community/

For OpenCode, copy skills from this repo into your project's .opencode/skills/ directory.

Default Stance

Each tool is documented in three privacy tiers:

Tier	Goal
Recommended	Keep normal foreground AI usage working while turning off telemetry and optional sharing where possible
Strict	Tighten local retention, sharing, and remote features when the tool supports it
Feature-max	Keep more hosted convenience features, but document the privacy tradeoffs explicitly

The checked-in config files are the Recommended profile unless the tool README says otherwise. OpenCode is the main exception: opencode/opencode.json is the generic multi-provider template, opencode/examples/openrouter-template.json is the generic OpenRouter-only template, and opencode/examples/routed-claude.json is a concrete Claude-via-OpenRouter example.

Two Different Privacy Surfaces

For every tool, this repo separates:

1. Local config: files like settings.json, config.toml, opencode.json, env vars
2. Vendor platform settings: web UI or admin settings such as training opt-outs, Bing/web search, repository access, memories, or data-sharing toggles

That distinction matters. Turning off local telemetry does not automatically disable:

• model training / product improvement
• server-side retention
• background-agent storage
• repository/account access scope
• provider-side logging under BYOK or third-party routing

Quick Install

Claude Code

cp claude-code/settings.json ~/.claude/settings.json

VS Code

# Linux/macOS
cp vscode/settings.json ~/.config/Code/User/settings.json

# Windows (from Git Bash)
cp vscode/settings.json "$APPDATA/Code/User/settings.json"

Cursor

# Linux
cp cursor/settings.json ~/.config/Cursor/User/settings.json

# macOS
cp cursor/settings.json ~/Library/Application\ Support/Cursor/User/settings.json

# Windows (from Git Bash)
cp cursor/settings.json "$APPDATA/Cursor/User/settings.json"

OpenCode

# Generic template: choose your own providers and model IDs
cp opencode/opencode.json <your-project-root>/opencode.json

# Generic OpenRouter-only template
cp opencode/examples/openrouter-template.json <your-project-root>/opencode.json

# Concrete Claude-via-OpenRouter example
cp opencode/examples/routed-claude.json <your-project-root>/opencode.json

Codex CLI

cp codex-cli/config.toml ~/.codex/config.toml

T3 Code

# Set before launching T3 Code (add to ~/.bashrc or ~/.zshrc for persistence)
export T3CODE_TELEMETRY_ENABLED=false

Zed

cp zed/settings.json ~/.config/zed/settings.json

Verification Model

The tool docs in this repo now use the following labels:

• Verified: confirmed in current official docs
• Dashboard-only: real control, but not managed through the local config file
• Best-effort: behavior is plausible or observed but not clearly guaranteed in current vendor docs

Use defaults-vs-recommended.md for the cross-tool matrix and setting-by-setting breakdown, and the per-tool READMEs for details.

Repo Structure

dev-tool-configs/
├── README.md
├── docs/
│   ├── README.md
│   ├── known-issues.md        # Tracked upstream vendor bugs affecting configs
│   └── skills-architecture.md # Canonical guide for where skills should live
├── defaults-vs-recommended.md
├── claude-code/               # Default config, annotated docs, and Claude skills
│   ├── skills/
│   └── skills-community/
├── codex-cli/                 # Default config, annotated docs, and shared Codex/OpenCode skills
│   ├── skills/
│   └── skills-community/
├── t3-code/                  # Env var config and annotated docs
├── vscode/
├── cursor/
├── opencode/                  # Default config, examples, docs, and native OpenCode-ready skills
│   ├── examples/
│   ├── skills/
│   └── skills-community/
└── zed/

Directory Guide

High-level path map:

Path	Purpose
claude-code/	Claude default config, annotated docs, and Claude-only skills
claude-code/skills/	Claude-specific curated generic skills
claude-code/skills-community/	Claude-only contributor favorites
codex-cli/	Codex default config, annotated docs, and shared skills source
codex-cli/skills/	Maintained source library for the shared Codex/OpenCode-compatible skill set
codex-cli/skills-community/	Shared contributor favorites, not part of the curated default set
t3-code/	T3 Code env var config and annotated docs
opencode/	OpenCode default template, routed examples, and native skill copies
opencode/examples/	Alternate provider and routing profiles
opencode/skills/	Native OpenCode-ready copy of the shared set plus any OpenCode-only variants
opencode/skills-community/	OpenCode community favorites and experiments
docs/skills-architecture.md	Canonical “where should this skill go?” guide

For exact placement rules, use the docs guide rather than the root README.

Maintenance Notes

• AI instruction files (CLAUDE.md, AGENTS.md, .cursorrules, .github/copilot-instructions.md) share the same project rules so that contributors get consistent guidance regardless of which AI tool they use. If you update one, update the others to match. CLAUDE.md is the canonical source.
• Shared skills (codex-cli/skills/ and opencode/skills/) are kept as separate copies so each tool directory is self-contained. codex-cli/skills/ is the maintained source. If you change a shared skill, copy the change to the matching opencode/skills/ file.

Disclaimer

These configurations are provided as-is on a best-effort basis for reference and convenience. They are community-maintained and are not guaranteed to prevent data loss, data collection, security breaches, privacy violations, model training, or server-side retention.

All settings in this repo are derived from vendor documentation available at the time of writing. Vendor defaults, setting names, and data practices can change at any time without notice or a corresponding update to this repo.

By using any configuration from this repo, you accept full responsibility for reviewing and verifying it before applying it to your own environment. Before using anything from this repo, you should:

• read the vendor's current official documentation
• verify the settings against your plan, account type, and tool version
• review vendor privacy policies and dashboard/account settings
• test configurations in a non-critical environment first where possible

This repo does not provide privacy, security, legal, or compliance advice. These configurations are not a substitute for regulatory compliance (GDPR, HIPAA, SOC 2, or any other framework). If you operate in a regulated environment, consult qualified professionals independently.

The maintainers and contributors accept no liability for any damages, data loss, or privacy incidents arising from the use of these configurations. See the MIT License for full legal terms.