Claude
Tools

ResearchX — AI Research Operating System. Literature-driven AI research assistant for full scientific lifecycle.

Multi-agent orchestration framework for Claude — tool registry, memory, parallel runners, orchestrator/worker topology

Companion cookbooks that take a provider's agent, swap only its read path for one NavFlow query, and measure the difference on a real running system. Includes a self-contained SRE incident-response benchmark — four injectable faults, provider-style tool fan-out vs. one correlated query

Patchistry MCP server — AI agent commerce tools (list canvases, list patches, recommend builds, shipping, contact). Built with Model Context Protocol SDK. Deployable to Vercel free tier.

Local bridge for the ChatPanel Chrome extension (Claude Code + Codex)

Pi extension to built with pocketflow dynamically workflows. Inspired on Anthropic's dynamic harness: https://claude.com/blog/a-harness-for-every-task-dynamic-workflows-in-claude-code

Chrome extension + local MCP server that lets any AI (Claude Code, Gemini CLI, or your own local LLM) drive your real browser session — no Playwright, no separate profile. Screenshots, page snapshots, clicking and typing with a visible cursor, plus a side-panel chat assistant for any OpenAI-compatible local model (Ollama, LM Studio, vLLM).

Symposium (συμπόσιον) — VS Code extension hosting dialogue sessions with multiple AI agent CLIs (Claude Code, Codex, Copilot)

The npm-audit for MCP servers: scan AI agent tools for dangerous permissions, prompt injection, and data exfiltration risks.

MCP server exposing Proxmox VE management as tools over stdio

VS Code extension that syncs AI setup files and MCP configs (Claude Code, Copilot, Cursor, Codex, Antigravity and more) from a GitHub repo into every project automatically.

Idiomatic Clojure wrapper over the official Anthropic Java SDK - messages, streaming, tools, batches, files, structured output

A Python CLI orchestrator that drives Claude (via the Anthropic API) through hypothesis-driven forensic investigations on the SIFT workstation. The LLM decides what to investigate and how to interpret results. The Python code controls what can be executed and what tools are available.

Agentic TypeScript pipeline · daily sales KPIs by code, narrative by Claude or Groq, delivered to Telegram, Slack and Email.

ChatPanel — Chrome/Edge side-panel chatbot for Claude Code, Codex, Gemini CLI & BYO models. Source-available (FSL-1.1).

Turn an OpenAPI/Swagger spec into a high-quality, curated MCP server — consolidates endpoints into clear tools and proves it with an eval harness.

MCP server giving Claude full Telegram tools (read/send/edit/pin/react/files)

VSCode extension that adds a virtual organization layer (folders, tags, links, and read-only smart groups) over Claude Code chat sessions, so 50+ chats stop overflowing the tab bar. Strictly read-only on Claude's transcripts.

A harness for testing, benchmarking, and tuning PostgreSQL extensions in an isolated, reproducible environment.

Enhance the tool-calling capabilities of your Claude Code, enabling it to deliver performance comparable to or even surpassing Anthropic Opus 4.7 when utilizing APIs of other models such as Qwen, DeepSeek, Kimi, and Gemma.（增强你的claude code的工具调用能力，让你使用qwen、deepseek、kimi、gemma等其它模型api时，能够获得接近甚至反超anthoropic opus 4.7的水平。）

A local-first agentic coding crew: a manager model plans, worker models build with real tools, and every subtask is gated on a runnable test — unverified work is never counted as passed. Local via Ollama; an optional Claude/Opus manager catches incomplete-test bugs.

Reach your own Claude Code from a Slack DM — local brain, your tools & memory, Socket Mode, single-owner. MIT.

Reusable patterns for personal agentic automations — YouTube monitoring, free transcription, CrewAI tools, GitHub Actions scheduling, and attributed-sentiment prompting

Burp Suite extension + port-based highlighter: dedupes HTTP history into a live unique-request feed and color-codes attacker/victim traffic by listener port (PwnFox-style) — built for Android/iOS multi-account IDOR/BOLA testing, with Magic Cookie, Match & Replace, and .http export for Claude Code / AI.