推理、Gateway、RAG、评测、可观测性与部署
平均值只计算有观测的数据；未知值不按 0 分计入。
Security for AI agents (built before the breaches started)
a deterministic gateway between AI agents and your systems
Deterministic-first execution engine for agent workflows in Go: the LLM extracts at the edge, a deterministic state machine decides. Zero dependencies, no-code JSON plugins, replayable runs.
In Security: OAuth 2.0, RBAC
SSD-streaming inference engine for giant MoE models (Rust + CUDA). GLM 5.2 743B at 2 tok/s and Hy3 295B at 7 tok/s on two consumer 16GB GPUs. Zero-config multi-GPU: measures PCIe bandwidth, places attention and hot experts where they fit.
Official DIDA Hotel Booking MCP Server. 14-year travel tech data stack, 2M+ hotels at wholesale rates, 40+ LLM compatible. Free unlimited calls for businesses & individual devs. Filter by location, date, star grade, guests & tags; pull real-time room types, pricing & cancellation rules.
Per-user cost dashboards from Claude Code's built-in OTel
Non-destructive compression gateway that cuts AI coding-agent token bills 25% on a single turn, climbing past 60% across long multi-turn sessions and past 85% in context-saturated deployments. Powered by the first open-source code-native 4B compression model. Drop-in for Claude Code, Cursor, Codex, OpenHands. No agent code changes.
Build your HUMAN.md.
A Mac vault that runs authenticated actions for AI agents over MCP. The agent gets the operation, never the key: no command reveals a stored credential, and there is no export route.
tune an 8B model on a 4 GB laptop GPU
Run MoE models bigger than your RAM. A 284B on a 12 GB phone, CPU only, lossless, on stock llama.cpp
Encrypted, fully offline agentic memory. One click install, GUI w/ memory map, all OS and agents. Superior memory creation, storage and retrieval.
MCP-based discovery and judgment layer for Korea’s open data portal—helping AI agents find, compare, and assess datasets with explicit evidence, JSON-LD/DCAT metadata, and versioned deterministic rules.
Multilingual CPU-only ASR with a 1.58-bit BitNet decoder
Local Inference, Faster
Agent multiplexer with map-charting capabilities
AWS-native knowledge graph RAG framework unifying two graph-retrieval methodologies — Microsoft GraphRAG community summarization and LightRAG dual-level keyword search — on Amazon Bedrock, Neptune, and OpenSearch. Multi-hop QA over document corpora, with incremental indexing and multilingual support.
An agentic browser driver for AI , few tools, full control, real stealth, god-tier token efficiency.
One Shared Memory for Claude Code, Cursor and Codex (MCP)
为纯文本模型"看图“设计更好的视觉工具箱和技能，支持多图理解，图片问答，前端UI还原、GUI 自动化等，并可选无缝接入多个主流agent，直接识别粘贴图片｜ A vision toolkit and skill designed for text-only llms — image Q&A, long-screenshot OCR, frontend UI restoration, and GUI automation, with optional seamless integration for Codex, Claude Code, Pi, Oh My Pi, and OpenCode
A 2.78-trillion-parameter Kimi K3 running inference on a single CPU in 8.24 GB of RAM. Portable C99: no BLAS, no framework, no GPU.
Graph-Orchestrated Agent Loop — a production-grade framework on LangGraph. Combine workflow graphs and agent loops, transpile Dify DSL to runnable code, swap wire protocols (Dify/OpenAI).
A production-oriented AI workflow runtime for building, validating, recovering, and shipping complex AI workflows as dependable services. 面向生产的 AI 工作流运行时：快速开发、验证和恢复复杂 AI 工作流，并将其稳定交付为服务。
RouterFuel is an async Rust LLM gateway routing to 330+ models — OpenAI, Anthropic, Google, Meta, Mistral, and every model on OpenRouter — with semantic caching and automatic failover.
one shared memory file for Claude Code, Codex, Cursor
Mindlas catches your coding agent drifting before the bad code lands. Real-time tracking and correction for context rot, unverified done claims, patch sprawl, and tool loops in your Claude Code session.
A WebRTC-native, audio-first conversational-AI framework for Go. - gojargo/jargo
cMCP: Confidential MCP Gateway. Hardware-attested policy enforcement for MCP tool calls. - agentrust-io/cmcp
Persistent, structured project memory for Claude Code and Codex - event-sourced, typed, fail-open. Your agent stops re-deciding what you already decided. - KanishkNoir/cognikernel
Generate 3D assets from coding agents
AI SRE AgenticOps for Kubernetes and cloud infrastructure.
RAG – RAG engine that documents what it doesn't do yet
Python SDK for LLM guardrails with safety classification, PII detection, prompt injection defense, and grounding checks. Protect AI applications locally with zero external APIs, streaming support, and plug-and-play integration for any LLM.
ZhuLing — Zero-config AI Agent framework for Java. One YAML to launch Agents on Spring AI + DDD, with MCP tools, full observability and a ReAct core. Ai Agent 快速开发框架
Open-source GraphRAG domain knowledge layer: LLMs continuously distill scattered skill docs into a queryable, evolvable, human-editable two-layer knowledge graph — versioned, hot-swapped and rolled back via a GitOps pipeline like software releases, and served to AI agents over MCP. RAG, but with relational structure.
Your agents already solved this. deja finds it — it indexes the sessions your coding agents already wrote to disk, months of history from before you installed it, and recalls them automatically at session start across seventeen harnesses. 84.9% hit@1 on LongMemEval-S, no LLM, no embeddings. One zero-dep binary, fully local.
A continuously-verified dataset of free-tier & trial-credit LLM APIs for developers. Every entry dated, sourced to the provider's own docs, and machine-readable. No hype, no dead links.
Extra Small Agent Framework
Context compression for MCP. Same upstream call, exact results recoverable.
compatible gateway for AI models
A token-spend profiler and cost-regression gate for AI agents.
a notes app. but also shared memory space for your agents (MCP)
Dotted thought-orb loading indicators for AI & agent UIs, 9 tuned types, two sizes, auto dark/light
Run large models like Qwen3.6-35B-A3B locally on a 16 GB RAM machine
The SQLite of agent sandboxes — self-hosted, E2B-compatible. One machine, sandboxes that live forever, idle costs nothing.
See what your coding agents did and what it cost. Breaks each task down into work steps — tools used, files changed, tests run, time and tokens spent. Local-first dashboard for Claude Code, Codex, OpenCode, and more. No login, no telemetry.
Persistent, provider-neutral memory for Codex, Claude Code, OpenCode, Pi, and MCP coding agents.
An egress gate for CLI coding agents. Masks secrets and customer data in the request body before it reaches the model provider, and the agent keeps working. - softcane/hamza
不死鸟 Phoenix — Hermes Agent 插件：路由分档/风险防线/自愈/存档点提醒/审批策略自适应，官方钩子接入，不改 Hermes 核心代码
Multi-hop RAG retrieval with zero LLM calls in the query path
a local runtime and hosting for LLM agents
Security control plane for AI agents — identity and delegation, capability policy, data-flow taint and a live audit trail, enforced over MCP. Guards a real Claude Code end to end.
TERSE - a hierarchical state language for humans and AI agents.
Search AI Agents from Any MCP Client
Research artifacts from Hyra (/ˈhaɪ.rɑː/)
GitHub repo stats: collect, analyze, TUI, agent inference
source control room for AI coding agents
Expanding AI Capabilities
🍙 A personal AI agent & local memory hub for all AI agents like Claude Code, Codex, OpenClaw and Hermes Agent. Gives every AI one shared, fully controlled memory and persistent context — all AI remember the same you.
source, Long-horizon cite-able memory for multi-agent systems
The bridge for remote MCP — seamless OAuth, resilient auth recovery, production-grade reliability
Open-source MCP connectors and skills for ChatGPT
Verifies model identities of your AI gateway
validate any MCP server compliance with a new MCP version
Credential proxy so AI agents never see the keys
ACID rollbacks and dry-run guardrails for AI agents
Cage untrusted MCP servers in containers, compose them into agents, and share them over any OCI registry. Signed, sandboxed, no Docker required. - okedeji/mcpvessel
Connect Claude Desktop, ChatGPT, Codex Desktop, OpenClaw, or any MCP-capable assistant to Revise. Agents can draft, edit, export, and share documents on your behalf — even before you have an account. Includes the full tool-by-tool reference.
Revenue-first website analytics installed and verified by AI agents through MCP
hosted AI gateway – MCP, budget, PII, smart router, fallback
Local-First Memory Across Claude Code, Codex, and Cursor
X-ray for documents: lossless PDF & PPTX extraction to JSON with bounding boxes, fonts, and colors — CLI, HTTP API, and an in-browser playground. Rust + PDFium.
Native macOS orchestrator for coding agents — run Claude Code, Codex, Cursor, Gemini and shells in parallel across git worktrees and remote hosts
an append-only log of what your AI agent already rejected
Poirot is a deep research agent kernel built for those who care about how agents are architected.
Let engineering agents author or import models, define materials, boundary conditions and loads, then simulate, verify and optimize through REST, MCP or Python.
Fail-closed reverse proxy and circuit breaker for AI agents
Markdown-defined provider-backed agents and deterministic workflows for ACP runtimes.
Codex skill for auditing and tuning Codex Desktop context/tool surfaces
MCP aggregator that exposes 2 tools instead of hundreds
Fast, pure-Python full-text indexing and search. Actively maintained continuation of Whoosh / whoosh-reloaded. Install: pip install whoosh3
Curated collection of modular agent skills for LLM-based agents
A programmable tool and agent runtime for Pi
git log for your infrastructure — tamper-evident host snapshots, diffable history, and drift rules for Linux (and your AI agent's config) - statedrift/statedrift
but-retrievable RAG vectors
DNA lineage test of Korean LLM & VLM foundation models
OpenShell Kubernetes Operator
coding agents collaborate across companies
Zero-modification Human-in-the-Loop adapter for OpenOPC's agentic DAG runtime. Park work items for human review. Resume execution automatically. No timeouts.
Governed execution cells for AI agents. Contribute to Runewardd/runeward development by creating an account on GitHub.
Self-hosted web viewer for Claude Code session transcripts — read, search, replay, and audit every session you have ever run
only 1 earned an A
see what compaction dropped from an agent session
a record format for AI evaluation runs, with reproducible digests
A Simple Probabilistic Program Inference Tool Supporting Loops - probabilistic-program-inference/README.md at master · kccqzy/probabilistic-program-inference
Shared Skill & Secret vault for your team and agents, without leaking keys. Join community here: https://discord.gg/6mQYYfFMAn