ThinkWatchProject

ThinkWatch Core — AI skill for Claude Code

AI community

Gateway engine for Claude Code, Codex and other AI clients, inside ThinkWatch Lite or on a Linux server: credential redaction, tool-call inspection, rule-based routing, per-request cost.

How to install ThinkWatch Core

This entry records only its repository, not the path inside it, so there is no exact command to give. Open ThinkWatchProject/ThinkWatch-Core and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What ThinkWatch Core does

Gateway engine for Claude Code, Codex and other AI clients, inside ThinkWatch Lite or on a Linux server: credential redaction, tool-call inspection, rule-based routing, per-request cost. Rust, MIT.

Alternatives in AI

  • Firecrawl MCP Server — 🔥 Official Firecrawl MCP Server - Adds powerful web scraping and search to Cursor, Claude and any other LLM c 6.1k ★
  • Phantom — An AI co-worker with its own computer 1.3k ★
  • AI Claude Start — Multi-profile Claude Code launcher with secure credential management 219 ★

README

ThinkWatch Core

Rust License: MIT macOS: arm64 Windows: x64, arm64 Linux: x86_64, aarch64

ThinkWatch Core

**[English](README.md) | [中文](README.zh-CN.md)**

ThinkWatch Core is the gateway engine behind ThinkWatch: a set of Rust crates and the `twcore` binary built from them. Claude Code, Codex and other clients of the Anthropic, OpenAI and Gemini APIs point at `twcore` once. From then on, credentials can be kept out of outgoing requests, the tool calls that come back are checked, and every request is recorded with where it went and what it cost. `twcore` runs inside the desktop app [ThinkWatch Lite](https://github.com/ThinkWatchProject/ThinkWatch-Lite) or on its own on a Linux server, and [ThinkWatch Enterprise](https://github.com/ThinkWatchProject/ThinkWatch) builds on four of its crates.

Documentation: [configuration reference](docs/config.md) · [running core on a server](docs/server.md) · [thinkwat.ch/core](https://thinkwat.ch/core/)

Highlights

  • Connect once, switch freely. Clients keep one address and one key; changing upstreams or models happens in the gateway, with no client change or restart. Anthropic Messages, OpenAI