Harness Bridge
SDK / ToolA local launcher that points Claude Code and seven other coding harnesses at any OpenAI-, Anthropic- or Responses-compatible endpoint — without editing a single config file.
- launcher
- local-models
- cli
- menu-bar
A local privacy gateway for AI coding assistants — masks credentials, connection strings and PII into structured placeholders on the way out, restores them mid-stream on the way back.
Data Maskit sits between your terminal and external AI services: requests are automatically masked into structured placeholders before they leave (sk-proj/ghp-style keys, cloud AccessKeys, JWTs, PEM private keys, database connection strings, private IPs, phone and ID numbers, internal codenames become {{CONNSTR_…}}, {{PHONE_…}} tokens), and the model's streamed reply is restored to the originals in milliseconds. The model works around placeholders and never sees real values. 100% local compute, zero telemetry. Ships as a desktop client (Windows, macOS Apple Silicon) and a Docker gateway (linux/amd64 and arm64) with a built-in web console — works with Cursor, Claude Code, Codex, Pi, OpenCode and other assistants. AGPL-3.0.
Teams and individuals who must keep credentials, internal topology and PII off third-party model providers — without giving up Cursor, Claude Code or Codex workflows.
# Docker gateway (linux/amd64 + arm64, built-in web console): docker run -d \ --name maskit \ --restart unless-stopped \ -p 127.0.0.1:5801:5801 \ -p 127.0.0.1:18701:18701 \ -v maskit_data:/data \ -e MASKIT_PANEL_TOKEN="YourSecretToken123456" \ ghcr.io/xiayutian11/maskit:latest # desktop client: grab the Windows installer or macOS dmg from GitHub Releases
AGPL-3.0. The panel token must be at least 16 ASCII characters or the engine safely falls back to a random token printed in the startup logs. Round-trip fidelity depends on rule coverage — test with your own secrets before trusting it blind.
Repository
xiaYuTian11/maskitCharted
A local launcher that points Claude Code and seven other coding harnesses at any OpenAI-, Anthropic- or Responses-compatible endpoint — without editing a single config file.
Answers macOS Spotlight and Siri prompts with your signed-in Claude Code account — an experimental macOS 27 model-delegation provider.
Extract your own local chat history from ten AI coding assistants into one normalized JSONL — for fine-tuning, analytics, or backing up years of conversations. Standard library only.