Explore

Recipe and scripts for online iterative RLHF: SFT, reward modeling, vLLM response generation, reward annotation, and iterative DPO training loops. Reproduces LLaMA3-8B alignment comparable to Llama3-8B-Instruct using only open-source data.

AI Frameworks | Python · rlhf · dpo

SMG is a Rust LLM gateway that unifies OpenAI/Anthropic/Gemini and self-hosted engines (vLLM, SGLang, TensorRT-LLM, TokenSpeed, MLX) behind one API. It adds KV-cache-aware routing, gRPC pipelines, MCP tooling, multi-tenancy and observability for large-scale inference deployments.

AI Frameworks | Rust · llm-gateway · inference-routing

AWS CDK construct library providing multi-service, well-architected patterns for generative AI on AWS: Bedrock, SageMaker model deployment, RAG knowledge bases, OpenSearch vector stores, agents, and batch inference. Use it to define repeatable GenAI infrastructure in TypeScript, Python, Java, Go, or C#. Experimental, n

AI Frameworks | TypeScript · aws-cdk · infrastructure-as-code

Governed shared memory layer for AI agent fleets: agents write plain text, Caura stores it with semantic search, entity/knowledge graphs, per-tenant isolation, trust tiers and audit trails. MCP-native plus REST and Python/TypeScript clients; self-hostable via Docker with Postgres/pgvector.

AI Agents | Python · agent-memory · multi-agent

MCP server and plugin that compresses AI coding agent context with cache-aware token optimization, enforces diff-only re-reads, and keeps a live per-project knowledge graph. Adds per-agent token accounting, a local savings dashboard, and benchmarks. Install via Claude Code plugin or npx; no telemetry.

MCP | JavaScript · mcp-server · token-optimization

Agent-agnostic SKILL.md pack that offloads small agent decisions (model routing, memory filtering, compaction, skill pick, triage, mail sorting, computer/browser action picks) to TypeSafe's Jev decision model. Ships a stdlib-only Python CLI, installer for Hermes/Claude Code/Codex, and a routing dashboard.

Skills | Python · agent skills · model routing

Official Aliyun Model Studio (Bailian/DashScope) CLI that exposes Qwen chat, image/video generation, speech, knowledge retrieval, fine-tuning and managed agents as structured tool calls for AI agents. Runs as an npm/curl install and ships MCP server and skill support.

AI Tools | TypeScript · cli · ai-agent

Self-hosted AI agent harness shipped as one Go binary: it plans multi-step tasks, executes sandboxed tools, and can write, test and repair new tools itself. Exposes those tools over MCP so Claude Code, Codex and other clients can share one tool library.

AI Agents | Go · mcp · mcp-server

A curated awesome-list of remote MCP servers you connect to over a URL via Streamable HTTP, with OAuth or API-key auth. Useful for discovering hosted MCP integrations for Claude, Cursor, and other clients without running local processes.

MCP | remote-mcp · awesome-list

Runtime refusal-direction ablation for Ternary Bonsai 2 27B: applies y -= alpha*(y·r)r at 129 residual writer sites in MLX, leaving the ~1.72-bit quantized weights bit-identical and alpha/layer tunable at inference.

AI Tools | Python · abliteration · uncensored-llm

Rust single-binary MCP server and CLI for building, changing, querying and reasoning over RDF/OWL knowledge graphs. It adds a native SHIQ reasoner, SHACL and SPARQL, plus proof-carrying inference checked by a separate Lean 4 verifier.

MCP | Rust · ontology · knowledge-graph

Claude Code plugin that turns the CLI into a business ops cockpit: 66 skills and 21 agents for infra, CI/CD, unified inbox, PR auto-merge, revenue, e-commerce, marketing and voice, plus a background daemon.

Skills | Shell · claude-code · claude-code-plugin

Blitz Strike is a TypeScript/Bun MCP server packaging a 3-tier pentest methodology: BLITZ recon, EAGLE-EYE source-to-sink taint analysis, and STRIKE live verification, with a 57-chain escalation graph and 130-tool catalog. Any MCP client can run full engagements via run_engagement.

MCP | TypeScript · mcp-server · pentest

A Claude Code skill and style-prompt library for directing and producing short films entirely with code. It includes 39 visual styles, production guides, and tools for frame-by-frame rendering, audio, and video assembly.

Skills | JavaScript · Claude Code · agent skill

A collection of 23 Python plugins that extend the hermes-agent autonomous agent with goal management, multi-model routing and debate, cost guardrails, telemetry, memory consolidation, and safety filters. Install by copying plugin folders into ~/.hermes/plugins/.

AI Agents | Python · plugins · autonomous-agent

macOS personal AI agent: an on-device LLM (Gemma via a LiteRT-LM fork) reads your files, messages and email nightly into a local markdown knowledge base, then drafts proactive morning actions. Adds a computer-use Sidekick and an opt-in zero-access-encrypted MCP server for ChatGPT/Claude.

AI Agents | Swift · on-device-llm · local-first

Official Anytype MCP server that converts Anytype's OpenAPI spec into MCP tools, letting AI assistants search spaces, create objects, lists, properties, and tags in a local encrypted knowledge base through natural language. Install via npx and add to Claude Desktop, Cursor, Claude Code, or LM Studio.

MCP | TypeScript · mcp-server · anytype

FreeCAD AI is an agentic assistant workbench for FreeCAD that turns natural-language prompts into 3D models by generating and executing Python, with Plan/Act modes, 50 tool-calling CAD operations, skills, MCP support and 21 LLM providers.

AI Agents | Python · freecad · cad

An MCP server implemented as a 1C:Enterprise extension, with an optional Python proxy for stdio transport and OAuth2 authentication. Includes built-in tools for 1C metadata and support for custom tools, resources, and prompts.

MCP | 1C Enterprise · Model Context Protocol · 1C:Enterprise

A local Python CLI that tracks token usage and estimated costs from Claude Code, Codex, and Kimi Code session data. It provides terminal reports, heatmaps, quota monitoring, statuslines, and a live session sidebar.

AI Tools | Python · token tracking · cost analysis

Local-first Electron desktop workbench for AI video: script, storyboard, image/video generation via any provider or local ComfyUI, timeline editing, and MP4 export. Ships 23 MCP tools so Claude Code, Codex, or Cursor can drive the whole pipeline.

AI Tools | TypeScript · local-first · electron

Nanocodex is a Rust library-first SDK that embeds OpenAI's Responses agent loop end-to-end: retained sessions, typed history, tools, Code Mode, branches, retries and cleanup. Node, Python, WASM and native CLI/desktop consumers prove the same contract, so you keep your own UI, memory and infrastructure.

AI Coding | Rust · openai · codex

Local-first codebase intelligence CLI and MCP server for AI coding agents. Builds a SQLite code graph across 28 languages, exposing 287 commands and 246 MCP tools for search, impact analysis, change-safety gates and audit evidence. No API keys or source upload.

MCP | Python · mcp-server · ai-coding

MCP server exposing Prometheus to AI agents: run PromQL instant and range queries, list metrics, read metric metadata and inspect scrape targets. Ships as Docker image and Helm chart with stdio/HTTP/SSE transports, auth, and multi-tenant support.

MCP | Python · mcp-server · prometheus

Python SDK for Stagehand’s AI browser automation API. It lets applications observe, act, extract data, and run multi-step browser tasks using natural-language instructions alongside Playwright and code.

AI Agents | Python · browser automation · computer use

Git-backed failure-memory library where AI agents search 393+ verified debugging lessons. Ships an MCP server with 7 tools (search, get_lesson, preflight, intake, write, register) over remote Streamable HTTP or local stdio, plus WebMCP, llms.txt and A2A discovery.

MCP | Python · mcp-server · failure-memory

Self-hosted backend for dispatching and monitoring AI coding-agent missions in isolated Linux workspaces. Supports Claude Code, OpenCode, Codex, Gemini, and Grok runtimes, with MCP control, project tracking, and a web dashboard.

AI Coding | Rust · AI coding agents · agent orchestration

OME is a Kubernetes operator for LLM serving: model-as-CRDs, GPU bin-packing scheduling, and automatic runtime selection across SGLang, vLLM, TensorRT-LLM and Triton. Adds prefill-decode disaggregation, multi-node inference, LoRA serving, autoscaling and benchmarking.

AI Frameworks | Go · kubernetes · llm-serving

tossctl is a Go CLI plus MCP server that exposes Toss Securities data (quotes, portfolio, dividends, taxes, order previews) to terminals and AI agents like Claude Code, Codex, and Cursor. Real trading is disabled by default and requires human-approved order previews.

MCP | Go · mcp-server · model-context-protocol

A native desktop AI app for chatting with hosted or local models, searching local files with RAG, connecting MCP servers, and delegating tasks to coding-agent CLIs. Stores conversations and RAG indexes locally.

AI Tools | Kotlin · desktop AI app · multi-provider chat

A Codex image-generation skill that turns uploaded images or text prompts into retro CRT-interface illustrations. It provides palette and style presets, aspect-ratio choices, and a Python script for final CRT effects.

Skills | Python · Codex skill · image generation

JIT-Agent is a meta-agent that generates and revises executable, task-specific harnesses for agentic LLMs. It includes harness designs, benchmark adapters, and candidate selection for evaluating workflows across research, office, planning, and other tasks.

AI Agents | Python · meta-agent · agent harnesses

1MCP is a unified MCP runtime that aggregates multiple MCP servers behind one `1mcp serve` process, with a CLI mode (instructions/inspect/run) for progressive tool discovery by agents. It supports stdio proxy, streamable HTTP, template servers, presets, filters and centralized auth.

MCP | TypeScript · mcp-gateway · mcp-proxy

An MCP server and companion agent skills for accessing freee business APIs. Supports accounting, HR, invoicing, and other freee services, with OAuth authentication, OpenAPI request validation, and tools for reading and modifying records.

MCP | TypeScript · MCP server · freee

Lunar.dev is a self-hosted API gateway plus MCP aggregator that sits between apps/agents and third-party APIs, adding traffic visibility, rate limiting, policy enforcement, retries and circuit breakers. MCPX consolidates multiple MCP servers behind one endpoint for security and observability.

MCP | TypeScript · mcp-gateway · mcp-aggregation

A collection of 900+ Markdown skills for autonomous AI research, organized into composable packages and a Campaign-to-SOP hierarchy. Supports literature discovery, gap analysis, hypothesis generation, experiment design, and session recovery, with MCP integrations for research and web sources.

Skills | research-agent · autonomous-research

A Go terminal UI for managing multiple AI coding agent CLIs in persistent tmux sessions. Track live status, send prompts, create worktrees, and review diffs with comments that return to the agent.

AI Coding | Go · AI coding agents · terminal UI

A self-hosted Rust control plane for coordinating parallel AI coding agents from a web dashboard or phone. It provides shared tasks, worker messaging, schedules, model switching, and session recovery for Claude Code, Codex, Gemini CLI, OpenCode, and Ollama.

AI Coding | Rust · coding agents · multi-agent orchestration

Open-source registry for MCP servers, AI agents, skills, and prompts, with a Go CLI (arctl), admin web UI, PostgreSQL backend, and Helm/Kubernetes deployment. It publishes, curates, and distributes artifacts and auto-configures AI clients.

AI Tools | Go · mcp · registry

Curated awesome-list of MCP servers for OSINT: SOCMINT, network scanning, web scraping, company registries, sanctions screening, threat intel, blockchain and geospatial. A directory to find and wire reconnaissance data sources into Claude, Cursor or other MCP clients — not a server itself.

MCP | osint · mcp-servers

MCP server exposing 122 Photoshop tools (atomic ops plus one-step recipes) so Claude, Cursor or a built-in chat UI can drive image editing in plain language. Ships a standalone web UI, ExtendScript/UXP bridge and Adobe Firefly generative features.

MCP | TypeScript · photoshop · adobe

Chinese-language, hands-on tutorial repo that teaches building AI agents end to end: LLM calls, function calling, LangChain, RAG with FAISS/Chroma, LangGraph workflows, MCP servers and clients, Agents SDK/Swarm multi-agent, and a Streamlit UI. 13 runnable modules, each with its own code and in-depth blog write-up.

AI Agents | Python · tutorial · langchain

Self-hostable platform for building and operating AI phone agents, with a LiveKit voice runtime, Twilio and Telnyx integrations, a customer console, call campaigns, knowledge retrieval, and MCP tool connections. Requires provider credentials for live calls.

AI Agents | TypeScript · voice agents · phone agents

Velorn is a cross-platform desktop video editor with generative workflows backed by ComfyUI and a local MCP server offering 100+ tools for agent-driven project review, timeline edits, generation, and export.

AI Tools | JavaScript · AI video · video editing

A personal AI chief-of-staff system that works through Claude Code or Cursor with a user-owned notes vault. It provides workflows for planning, meeting preparation, task follow-up, and personal knowledge management, with optional connected services.

AI Agents | Python · AI assistant · chief of staff

OpenAI's customer-run daemon that connects private or localhost MCP servers to ChatGPT, Codex, the Responses API, and AgentKit via an OpenAI-hosted tunnel, avoiding public exposure. Go CLI with health/metrics endpoints, OAuth support, and an embeddable Go SDK.

MCP | Go · mcp-tunnel · mcp-gateway

MCP server for the Hevy fitness app exposing 22 tools to read, analyze, create, and update workouts, routines, exercise templates, and body measurements. Works with hosted Streamable HTTP OAuth or local stdio via npx/Docker; requires a Hevy PRO API key.

MCP | TypeScript · mcp-server · model-context-protocol

Appium MCP is a TypeScript MCP server that gives AI assistants tools to drive Android and iOS automation: session management, vision-based element finding, locator generation, and Java/TestNG test code generation. Works with Cursor, Claude Code, and Gemini CLI.

MCP | TypeScript · mcp-server · appium

Rust CLI that indexes past AI coding agent sessions (Claude Code, Codex, pi, Hermes) into a local Lance dataset, then serves hybrid vector+BM25 recall, MCP tools, and Hugging Face Hub publishing of the memory.

RAG | Rust · agent-memory · session-indexing

Skill pack that teaches Hermes Agent to write its own analytical 'prism' prompts before analyzing an artifact, then report findings with line refs plus an explicit constraint report of what it missed. Seven proven prisms also work alone as system prompts.

Skills | PowerShell · agent-skills · prompt-engineering

Local-first control plane that runs Codex CLI, Claude Code, Cursor CLI and OpenCode behind one typed interface, with quota-aware multi-account rotation, best-of-N races, cross-model review and auditable cost/evidence receipts. CLI + daemon, macOS app, MCP server.

AI Coding | TypeScript · coding-agents · multi-harness

OpenCode plugin that turns one AI coding session into an architect-led agent swarm: specialized planning, coding, review, test, and docs roles with gated pipelines, scope enforcement, shell-write guardrails, and resumable .swarm state.

AI Coding | TypeScript · multi-agent · agent-swarm

A local OpenAI-compatible gateway that ranks free models across configurable providers and fails over when a provider is unavailable, rate-limited, or returns no results. Includes a web UI for provider configuration and usage monitoring.

AI Tools | JavaScript · LLM gateway · model routing

A curated catalog of agent-native services and infrastructure, with selection criteria and onboarding guides. Includes installable skills that help agents find, evaluate, and connect to services across categories such as email, browsers, memory, sandboxes, and MCP tools.

AI Agents | Python · agent infrastructure · service directory

Rust CLI and MCP server that indexes a codebase into embeddings and a tree-sitter AST knowledge graph, exposing semantic search, GraphRAG navigation, structural search, and optional LSP tools to Claude, Cursor, and other MCP clients. Local-first, Apache-2.0, with a reproducible retrieval benchmark.

MCP | Rust · mcp-server · semantic-search

Desktop/CLI app that turns LLM conversations into an editable thought graph where wires define the model's next context. Includes a read-only MCP server to search local Claude Code, Codex, DeepSeek Harness and Pi sessions.

AI Tools | TypeScript · infinite-canvas · thought-graph

GrayMatter is a local-first Go memory layer for AI agents, available as an MCP server, CLI, and Go library. It stores and retrieves relevant facts across sessions, builds an optional knowledge graph, and integrates with clients such as Claude Code, Cursor, and Codex.

MCP | Go · agent memory · persistent memory

CoCo adds agent orchestration, 226 skills, 386 commands, specialized agents, and persistent project state to Claude Code, Cursor, and Codex. It provides workflow structure and prompts; execution runs through the host coding assistant.

AI Coding | HTML · agent orchestration · agent skills

Local-first Context Assurance layer that selects budgeted, query-relevant context for AI agents and compresses it reversibly. Every reduction ships an auditable receipt plus a handle to recover the exact original bytes. Integrates via MCP, proxy, SDK, or agent wrapper.

AI Tools | Python · context-compression · token-optimization

A TypeScript MCP server and CLI that exposes 18 Google Maps tools for AI agents, including geocoding, place search, directions, route planning, weather, and air quality. Supports stdio and Streamable HTTP transports, plus an agent skill for CLI use.

MCP | TypeScript · Google Maps · geospatial

Rust retrieval engine with TypeScript and Python SDKs that helps AI agents discover relevant tools, skills, and facts on demand. Uses BM25 by default, with optional semantic and hybrid search, to reduce context and avoid loading every capability up front.

AI Agents | Rust · context engineering · tool selection

ICLR 2026 research benchmark for evaluating memory in LLM agents across four competencies: accurate retrieval, test-time learning, long-range understanding, and conflict resolution. Ships harnesses for long-context, RAG, and agentic memory methods (mem0, Zep, Letta, Cognee) plus LLM-as-judge scripts.

AI Tools | Python · agent-memory · benchmark

CoWork OS is an MIT-licensed, local-first agentic workspace (Electron GUI plus cowork CLI) that runs coding, knowledge-work, browser and document tasks across bring-your-own model providers, with MCP connectors, skills, channels, memory and governed approval profiles.

AI Agents | TypeScript · agentic-os · local-first

Self-hosted podcast server that transcribes episodes with Whisper, uses an LLM to identify ad segments, and serves edited audio through RSS feeds. Supports Anthropic, Ollama, OpenRouter, and OpenAI-compatible providers.

AI Tools | Python · podcasts · ad detection

Mingyu is an open-source Chinese metaphysics toolkit that computes Bazi, Ziwei, tarot, Liuyao, Qimen and other charts and emits structured prompts for LLM interpretation. Ships an MCP server (remote SSE + npx), an agent skill, REST API and the mingyu-core npm package.

MCP | TypeScript · bazi · ziwei

MCP server wrapping TypeSafe's Jev (System One) model as three typed judgment tools: verify claims against evidence, screen text for prompt injection, and rank candidates semantically with no embeddings. Returns calibrated probability distributions in ~150-500ms.

MCP | JavaScript · model-context-protocol · typescript

An MCP server that lets AI assistants automate Firefox through WebDriver BiDi. It provides tools for page navigation and interaction, screenshots, console and network inspection, and browser debugging; it requires a local Firefox installation.

MCP | TypeScript · Firefox · browser automation

A browser skill that delegates navigation, clicks, toggles, and scrolling to Jev while Codex handles text entry, visual judgment, and verification. Uses an existing Codex Computer Use connection to Chrome or its in-app browser.

Skills | JavaScript · browser automation · computer use

A TypeScript template for a lead agent that delegates marketing work to five specialists for content, social, SEO, product messaging, and email. Agents work with Notion, Typefully, and Resend, with human approval required for consequential actions.

AI Agents | TypeScript · multi-agent · marketing

Python toolkit from AWS Labs for GraphRAG: builds hierarchical lexical graphs from unstructured data and composes graph-based question-answering strategies. Includes BYOKG-RAG for KGQA over your own knowledge graph, with integrations for Neptune, OpenSearch, PostgreSQL and LlamaIndex.

RAG | Python · graphrag · knowledge-graph

Curated awesome list of RAG research: surveys, papers, lectures, workshops and tools, with Semantic Scholar citation badges for each entry. Use it as a reading map for retrieval-augmented generation techniques and architectures.

RAG | awesome-list · retrieval-augmented-generation

Koishi chatbot plugin that adds multi-model LLM chat (OpenAI, Claude, Gemini, DeepSeek, Qwen, Ollama and more) via a LangChain-based adapter layer. Offers chat/browse/agent modes, YAML persona presets, MCP client tools, long-term memory, web search and text, voice or image output.

AI Frameworks | TypeScript · chatbot · koishi

A collection of Jev use cases, workflows, and installable agent skills for coding agents, shipping 5 skills, 108 scenarios, and an eval suite covering calibration, model panels, red-teaming, and skill-gating. Useful if you want ready-made skill packs plus reproducible evaluation tooling.

Skills | Python · agent-skills · coding-agents

Community-maintained directory of 3,000+ bug bounty and vulnerability disclosure programs, shipped as an MCP server, REST API and web app. The TypeScript MCP server lets agents search programs, policies and disclosure contacts for security research.

MCP | TypeScript · bug-bounty · security

Late is a Go CLI coding agent that assigns planning and verification to an orchestrator while isolated coder and researcher subagents handle execution. Its context-isolation design supports long workflows, local or cloud models, MCP, plugins, and sandboxed runs.

AI Coding | Go · coding-agent · multi-agent

ai4j is a JDK 8+ Java agentic SDK giving one API across OpenAI, Anthropic, DashScope, DeepSeek, Ollama and more, plus Tool Calling, MCP, A2A, RAG, Agent Runtime and a built-in Coding Agent CLI/TUI/ACP.

AI Frameworks | HTML · java · jdk8

mcpstore is a Rust-built manager for MCP services with CLI, desktop GUI, HTTP API, and Python/Rust SDKs. It orchestrates MCP server lifecycles, groups tools per agent, and converts MCP tools into LangChain, LangGraph, OpenAI, AutoGen, CrewAI, or LlamaIndex tools.

MCP | Rust · mcp-hub · mcp-client

RedisVL is the AI-native Python client for using Redis as a vector database. It covers index/schema management, vector, hybrid and filtered search, embedders, rerankers, semantic caching, LLM memory, semantic routing, plus a bundled MCP server.

RAG | Python · vector-search · redis

Rust-native CLI and MCP server that builds a persistent SQLite map of a repository (folders, files, purposes, symbols, relations) so coding agents like Codex, Claude Code, and OpenCode locate the right code before reading files, with local token-savings telemetry. Worktree-aware and fully local, no hosted index.

MCP | Rust · code-intelligence · repository-map

A local-first desktop app that converts PDFs, Office files, scans, and other formats into structured Markdown for LLM and RAG workflows. Includes OCR, batch processing, and optional AI cleanup; conversions run offline.

AI Tools | Python · document conversion · Markdown

Rust-based linter and language server that validates AI coding assistant configs: CLAUDE.md, AGENTS.md, SKILL.md, hooks, plugins and MCP files. Ships 456 rules with autofixes, a CLI, GitHub Action and LSP plugins for VS Code, JetBrains, Neovim and Zed.

AI Coding | Rust · linter · lsp

Memento MCP is a TypeScript MCP server that gives LLM clients (Claude Desktop, Cursor, Copilot) persistent long-term memory backed by Neo4j. It offers a versioned knowledge graph of entities and relations with vector semantic search, point-in-time history, and time-based confidence decay.

MCP | TypeScript · knowledge-graph · memory

A Rust runtime and server for pulling and running decision models locally, including classifiers and NLI models. Provides an Ollama-style CLI, a TypeSafe-compatible API, and an MCP server for agent access.

AI Tools | Rust · local inference · decision models

A FoJin-powered Buddhism AI persona framework shipped as Anthropic-format Agent Skills for Claude Code, Cursor, Codex, OpenCode and Gemini CLI. It provides 15 prebuilt master personas plus compare/debate/curriculum meta-skills, with source-grounded citations, HARD-GATE ethics rules and automated fidelity/citation-audit

Skills | Python · agent-skills · ai-persona

Interceptor gives AI agents browser, macOS, and physical iPhone automation through a CLI and MCP server. It supports structured UI inspection and actions, workflow recording, and verification using the user's existing browser sessions.

AI Tools | TypeScript · computer-use · browser-automation

Smithers is a TypeScript agentic workflow framework: you declare flows next to your code and it runs coding agents through plan, execute and review steps with durable execution and human-in-the-loop gates. Ships a CLI, libraries and a web UI for automating repo changes.

AI Agents | TypeScript · agent-orchestration · multi-agent

CAIPE is an open-source platform for building, deploying, and governing AI agents and agentic workflows. It combines multi-agent orchestration (LangGraph/DeepAgents), MCP tool integrations, agent skills, identity/authorization, and observability, targeting platform engineering and enterprise DevOps use cases.

AI Agents | TypeScript · multi-agent-systems · agent-orchestration

Pulse is a macOS menu-bar/edge-docked monitor showing remaining quota, rate-limit windows and reset times for 20 AI coding services (Claude Code, Codex, Cursor, Copilot, Grok, Kimi and more) using your existing logins. It also offers optional local token-spend tracking and JSON output for tmux/sketchybar/Raycast.

AI Tools | Swift · macos · menu-bar

Docs build pipeline and MDX source for docs.langchain.com, covering LangChain, LangGraph, LangSmith, and Deep Agents. It is a documentation monorepo, not an AI library or agent you install.

AI Frameworks | MDX · documentation · langchain

Zero-shot text classifier served by a single Cloudflare Worker: plain HTTP GET, CLI (classifier-dev), and MCP server, no API key or account. Batches up to 1000 texts per request with calibrated probabilities, multi-label support, and an LLM fallback chain when the primary decision model is down.

AI Tools | TypeScript · zero-shot-classification · text-classification

REA is a CLI and MCP server that lets coding agents inspect applications, JavaScript behavior, and native binaries using tools such as Hopper or Ghidra. It returns structured evidence and investigation results to help explain features and guide reimplementation.

AI Agents | TypeScript · reverse-engineering · binary-analysis

Hermes Agent plugin adding web_search_plus and web_extract_plus tools that return real source links and cleaned page text across 15 search and 9 extraction providers, with fallback routing, caching, and an optional local key-free DonSeTch MCP provider.

AI Tools | Python · web-search · web-extraction

Cross-platform security terminal and vulnerability scanner with autonomous agents for audit, reconnaissance, CTF, and CVE workflows. Includes SBOM analysis, multi-protocol remote access, and an AI assistant supporting DeepSeek, OpenAI, and custom APIs.

AI Agents | HTML · cybersecurity · security agents

Open-source library of 1,800+ accountant-reviewed tax guides across 230+ jurisdictions, served to LLMs via an included Python MCP server, agent skills, and per-country packages. Each guide is either reviewed by a named licensed accountant or source-cited, with nightly accuracy checks.

MCP | Python · tax · accounting

Curated awesome list of LLM pre-training resources: technical reports from Llama, Qwen, DeepSeek and others, training frameworks (Megatron-LM, DeepEP, DeepGEMM), open datasets and data-filtering methods. Useful as a reading and reference index for anyone pre-training or studying LLM training pipelines.

AI Frameworks | awesome-list · llm-pretraining

MARM Memory is a local-first MCP server that gives AI coding agents persistent memory in SQLite. It fuses session history, repository code graphs and concept graphs behind 14 MCP tools over HTTP/STDIO, with semantic re-ranking and no cloud dependency.

MCP | Python · mcp-server · agent-memory

Jev Search is a Cloudflare Workers web app that sends your query to TypeSafe's Jev model, which picks sources, time range and search terms, then ranks Search1API results with visible relevance scores. It streams results per engine as NDJSON and returns links and snippets only, no generated answers.

AI Tools | TypeScript · web-search · llm-ranking

A collection of Agent Skills for A-share quantitative trading, covering financial data access, strategy development, backtesting, trading APIs, and investment research. Skills can be installed in Claude Code, Codex CLI, and other compatible agents.

Skills | Python · agent-skills · quantitative-finance

deco Studio is a self-hostable control plane for AI agents: one governed MCP endpoint with an encrypted token vault, agent publishing, model routing (OpenRouter or direct providers), SSO/RBAC, audit logs, and per-thread token and cost observability. Runs locally via bunx or self-hosted.

MCP | TypeScript · mcp-server · mcp-client

A TypeScript MCP server and CLI that builds a fully local hybrid (semantic + keyword) index over PDF, DOCX, Markdown, text, and HTML documents. Runs embeddings and vector storage on-device with LanceDB and Hugging Face models, so private docs stay searchable offline from Claude Code, Cursor, Codex, or the terminal.

MCP | TypeScript · mcp-server · local-rag