npm agents

Give any LLM agent the power to read any website — even the ones that block bots. An MCP server that returns clean Markdown and handles blocks automatically via residential proxies.

Deterministic behavioral audit harness for LLM agent transcripts — structural signature, rigidity trajectory, and sandbagging/eval-gaming detection. No model access, no LLM calls, reproducible.

Circuit SDK agent runtime — write an autonomous agent (CircuitAgent) with pluggable custody (paper · self-custody · off-box signer · non-custodial vault). Buy/sell only; on the mesh the key stays off-box, so funds can't leave.

基于大语言模型 (LLM Agent) 和现代排版框架构建的 ACM/ICPC 自动化出题工作流。

Context deduplication for LLM agent sessions: repeated tool output is replaced with a short ref token

LLM agent + LLM-as-Judge + Human-in-the-Loop review + learning loop, as an embeddable library

Lightweight LLM agent framework built on node-llama-cpp — Agent, Memory, and Tool abstractions for local llama.cpp models

Model Context Protocol server for Posecode: let any LLM agent validate, ROM-check, and get a render link for a .posecode movement, natively.

Neuralis — self-hosted, multi-LLM agent platform. Name reserved; `npm create neuralis` scaffolds the community host. See https://neuralisapp.com

MCP server giving any LLM agent live stock quotes, crypto prices, SEC filings, XBRL financials, FX rates and economic indicators — no API keys required.

AI agent skill for generating dark-themed architecture diagrams as self-contained HTML files. Works with Claude Code, Cursor, Windsurf, and any LLM agent.

MCP server exposing server-computed planetary positions (Swiss Ephemeris), validated against JPL Horizons (NASA). Plug any LLM/agent into the real sky: 19 astrology tools, never hallucinated.

npm-installable CLI wrapper for agent-trace, deterministic record and replay for LLM agent HTTP traffic. Forwards straight to the real Python agent-trace CLI. Requires Python 3.10+ and the agent-observability-trace-cli PyPI package.

本地真实 LLM Agent 通过 WebSocket 连接 Worker/MatchDO。CLI 不持有本地引擎， 撮合、座位、投影视图和动作提交全部走生产协议。

Typed content blocks and message attribution for LLM agent output — one shape for chat transcripts and documents.

Human-in-the-loop approval gate for LLM agent tool calls. Classify tools read/write, block destructive calls behind an approver, and never persist a grant across turns.

Static analysis CLI that finds deterministic LLM-agent cost-exposure bugs (unbounded retry loops, unbounded while(true) agents, leaked system prompts) and computes a worst-case dollar exposure ceiling from real, dated model pricing.

optimus — meta-skill for LLM/agent tooling (wiki, caveman, aura, graphify). Installs the prebuilt Rust binary as `optimus`.

MCP server that gives an LLM agent a funded, spending-limited, private Sirius account — balance, private transfers, x402 paid fetch, quotes and swaps, over stdio

Gotong LLM agent base — provider interface and LlmAgent for building LLM-backed participants.

Тонкий шар роботи з LLM (локальні omlx + хмарні провайдери) поверх pi: model tiers, one-shot, agentic-раннери, write-guard, trace, telemetry, prompt-budget

Pointer to the Rulvar umbrella package: install @rulvar/rulvar (or this alias, which re-exports it). An embeddable TypeScript engine for durable, budget-bounded, testable multi-agent LLM workflows.

Leaving Blank Space Agent — 多 Agent 编排框架

本地真实 LLM Agent 通过 WebSocket 连接 Worker/MatchDO。CLI 不持有本地引擎， 撮合、座位、投影视图和动作提交全部走生产协议。

Opt-in OpenTelemetry (OTLP) trace + metrics publisher for the Agent Architecture Board. Streams LLM generations and container tool spans to any OTLP/HTTP backend — the workerd-safe fetch exporter on the Cloudflare Worker facade, the official @opentelemetr

Dark Harness (dh) — a single-binary LLM agent harness for autonomous dark-factory software work, with console and web UIs. Name reservation placeholder; a real multi-platform release is coming (see https://github.com/stefanrusek/dark-harness).

Governance for AI-written project docs: verify, catch drift, keep them code-grounded, enforce in CI. OKF-compatible, zero-dependency.

剧本阶段统一 CLI：素材转 Script v4，并兼容 Script v3 的本地写作、精修、校验和发布。

Multi-LLM MCP server for Claude Code & Codex — route AI agent work across 20+ low-cost provider buckets with round-robin dispatch and quota-aware fallback.

Provider-neutral model adapter layer: canonical messages, capabilities, streaming contract

LLM capture proxy with Anthropic cache_control TTL stamping, vision handoff, concurrency gating, rate limiting, and a live inspection dashboard

An open-source WebMCP-powered in-page GUI agent. Let a natural-language agent live inside your web app and drive its UI — no script injection, no browser extension, no headless browser.

AI-powered analytics dashboard generator TUI — https://openboard-site.vercel.app

AI agent SDK for web pages — embed a chat assistant that edits page data via schema-validated tools. A lighter, framework-agnostic alternative to CopilotKit/LangChain for in-page JSON-editing agents. Vue-bundled; works with DeepSeek, OpenAI, MCP.

General-purpose agent with transport abstraction, state management, and attachment support

Rulvar core: L0 contracts, journal kernel, ctx primitives, agent runtime, model router, tool system, dynamic orchestrator, InMemory and JSONL stores, event stream.

General-purpose agent with transport abstraction, state management, and attachment support

Local AI gateway for OpenCode with tool/function calling — use any model via OpenAI, Anthropic, or Gemini API format

Neuralis — self-hosted, multi-LLM agent platform where AI agents are coworkers, not chat windows. Name reserved; the community host installer ships here. See https://neuralisapp.com

LLM integration plugin for PostGraphile v5 — server-side text-to-vector embedding and text companion fields for pgvector columns

Downcity Agent 运行时 — 单 Agent 执行壳与本机 RPC 能力

LLM-judged CLI test runner — define test cases in YAML, execute commands, let an LLM judge PASS/FAIL. Outputs JSONL trace.

Capix Code — decentralized AI coding agent with GPU marketplace

Deterministic local eval utilities for AgentInspect traces

Passively detects when agents make follow-up promises and automatically delivers on them. Cron is for things you plan. Oath Keeper is for things you promise.

Rulvar evals: eval cases, golden outputs, rubric and judge graders, matrix sweeps, canary fingerprint.

General-purpose agent with transport abstraction, state management, and attachment support

面向 Ubuntu 的安全检测/加固/防护 skill，供 Claude Code / Codex 等 AI agent 调用，支持 npm/npx。命令：scan / run / status / harden / prune / fw-log / fw-rules / fw-optimize / fw-plan / fw-monitor / fw-verify / sensor-setup / sensor-collect / sensor-status / sensor-tune / correla

sara-agent - AI Coding Agent

General-purpose agent with transport abstraction, state management, and attachment support

Rulvar first-class provider adapter for the OpenAI Responses API, plus the openaiCompatible factory.

Rulvar test harness: createTestEngine, FakeAdapter, VCR cassettes, replay-strict runs, matchers.

Agent loop, tool runtime, event stream, context strategy, limits and cancellation

General-purpose agent with transport abstraction, state management, and attachment support

Persistent self-improving agent memory — install & forget: auto-inject context via git hooks, zero-arg bootstrap, machine-wide global vault, 13 memory types, confidence tracking, XML codebase graphs

Compile messy document directories into Karpathy-style llm_wiki knowledge bases — standalone Q&A CLI + Claude Code/Codex skills

Mastra-backed AgentHarness implementation for @salesforce/sfdx-agent-sdk

Scaffold and manage an LLM-maintained personal wiki

General-purpose agent with transport abstraction, state management, and attachment support

Rulvar determinism lint rules with structural JSON diagnostics for the planner self-repair loop.

Rulvar shell: run/resume/runs/inspect/plan/kb commands, TUI progress, createServer, createWorker, OTel exporter.

Lightweight runtime library for brt-built botruntime agents

Drop-in adapters that expose the Circuit Data API (pay-per-call in CIRC via x402) to agent frameworks: an ElizaOS plugin and a Solana Agent Kit action set. One dependency turns any agent into a Circuit data consumer.

Eval harness for Last Light workflows — drives the real production workflows against a mocked GitHub and grades deterministically (SWE-bench compatible).

Тонкий шар роботи з LLM (локальні omlx + хмарні провайдери) поверх pi: model tiers, one-shot, agentic-раннери, write-guard, trace, telemetry, prompt-budget

Sema — your own Claude Code-grade coding agent, in the terminal.

Rulvar PostgreSQL store implementing JournalStore and LeasableStore with a fencing epoch, for multi-process and multi-host deployments.

An opinionated Pi agent configuration. Includes: custom theme, plan mode, artifacts, subagents, LLM council, configurable restrictions, MCP support and a styled UI.

General-purpose agent with transport abstraction, state management, and attachment support

Manage human-curated context assets for LLMs and agents in Git.

Probe any ACP-compatible agent for its capabilities (models, modes, configOptions, prompt capabilities, auth methods, MCP transports). No real prompt required; no token cost.

Rulvar first-class provider adapter over @anthropic-ai/sdk.

Professional command-line tool for HubDoc document management and bulk import/export

A TypeScript framework for building AI agent-powered applications. Integrates with multiple LLM providers (OpenAI, Anthropic, Ollama, Fireworks) and implements the AG-UI protocol for real-time frontend communication.

Claude Agent SDK-backed AgentHarness implementation for @salesforce/sfdx-agent-sdk

Keep your LLM wiki honest: a per-claim citation lockfile, drift detection, and code-grounding for Karpathy-style knowledge bases. Scaffolds a fresh wiki or adopts an existing one; agent-agnostic.

Rulvar adaptive orchestration extension: PlanRunner, RunLedger, escalation extensions, ModelLadder configuration. Replans during the run; not the plan-writing hybrid mode, which is @rulvar/planner.

Native cross-platform MCP agent. Inspector → MCPAgent → loop → fetch + @mcp-use/client. LangChain bridge at @mcp-use/agent/langchain.

A drop-in spend cap and audit log for LLM API calls

정크 컨텍스트 제거 — LLM 에이전트를 위한 보이지 않는 정보 정돈 레이어

Rulvar bridge adapter wrapping any Vercel AI SDK LanguageModelV4 as a ProviderAdapter.

Rulvar executable store conformance kit (DEF-4).

Rulvar SQLite store implementing JournalStore and LeasableStore with a fencing epoch.

Node HTTP broker for CommandSuite — the self-hostable command layer for agent teams (Hono + SQLite, one process).

Encode JSON → TOON (40-60% token savings for LLMs!)

NexusAgent - A powerful LLM Code Agent with MCP, Skills, and ACP support

Glassray Coach — a fully self-contained local AI-agent trace debugger (OTLP in, dashboard + agent-first CLI out).

Terminal agent that turns Lark/Feishu content into helios-kanban tasks

Zero Npm Dependencies Agent Framework

Reference agent that consumes the Knowledge Context Protocol end-to-end — a deterministic, fail-closed navigation planner plus optional LLM synthesis.

Framework-agnostic engine for agentic browsing: DOM accessibility snapshotting, element registry, action execution, tools, guardrails, and the LLM agent loop.

WAGENT Core - Gateway, Agent Engine, Memory

Rulvar flagship hybrid mode: plan agent, compileScript, WorkerSandboxRunner, self-repair loop. Plans before the run; not the PlanRunner orchestration extension, which is @rulvar/plan.

Gateway Agnostic SDK for Punk's hosted AI gateway: OpenAI, Anthropic, OpenRouter, Vercel AI SDK, LangChain, agent tracing, tool caching, and governance.

Turn LangChain agent runs into portable Zansn evidence packs. Capture traces, tool calls, cost, and readiness with a callback handler.

LangChain tool wrappers for the CRYPTYX x402 API. Drop-in DynamicStructuredTool instances that any LangChain agent can bind — signals, factor scores, institutional trigger evaluation, natural-language query. Payment plumbing handled via @cryptyx/x402-clie

Framework-agnostic Agent Protocol engine for LangGraph.js — run engine, handlers, and SSE, driven entirely by injected dependencies.

Framework-agnostic Agent Protocol engine for LangGraph.js — the heart of skein-js.

A live canvas for LangChain agents — stream documents, charts, and rich artifacts into a React panel.

langgraph.json parser and graph loader (path:export) for skein-js.