"A tool for searching local files" matching MCP servers:

Enables AI agent hosts to exchange durable, local sender/receiver messages through a SQLite mailbox exposed as MCP tools (connect, send, receive, reply, ack, status, disconnect), so a coordinator can dispatch commands and an executor can return results without model polling while idle. A configured host wake adapter is still required to turn an incoming event into an Agent turn.

A persistent local MCP relay that caches tool discovery, keeps upstream servers warm, and allows agents to hot-add or refresh MCP servers through management tools.

前端ai主导的全流程，自动化的预开发；旨在解决前端开发在联调前期的各种繁琐工作

Enables AI agents to control a Playwright-driven browser through natural-language instructions, including page interaction and live browser task execution.

Lets a human review what an AI agent is about to post to a chat tool (Slack first) before it is sent. Renders the draft, destination and warnings (@here/@channel, unresolved mentions, format issues) in a neutral local HTML preview; never sends anything.

A configurable MCP gateway that runs multiple Streamable HTTP MCP servers and exposes all their tools through a single endpoint, enabling tool aggregation and routing for MCP clients.

Enables delegating mechanical or high-volume subtasks to a local LLM through an MCP tool, letting the assistant query the local model without using its own output tokens for content.

基于MCP架构的微信公众号文章爬虫系统，支持AI智能体通过Selenium自动抓取微信公众号文章内容和图片，实现文章内容的智能分析。

Plori runs AI agents in Workspaces. A Workspace is a project folder in the cloud. Several agents can work in it at the same time, each on its own copy of the files, and Plori merges their work back into the folder. Every version of the files is kept.

Enables a local LLM to ask the user for guidance via a native desktop pop-up dialog, with configurable options and multiple-choice or free-text input.

Composes arbitrary MCP tools using JavaScript, enabling discovery, composition, and execution of other MCP servers through a single 'exec' tool.

Routed is a universal local router for Agent Skills across AI coding environments.

Enables delegating tasks to a separate, full agentic loop running on a different model, with its own tool access, returning only the final result.

Enables real-time user intervention for MCP agents through a web UI, allowing users to review context and provide feedback when AI agents drift from intent, keeping them on track.

Tells an agent what went wrong with its own Ollama calls, in sentences with the fix: lost prompt cache and where the prompt stopped matching, context overflow, a model running partly on CPU, tool schemas crowding the prompt. Also the last turn in numbers and which local models fit in video memory. Local, read-only, nothing leaves the machine.

Manages and dispatches multiple MCP servers from a cloud dashboard, reducing tool-context bloat by loading only the most relevant servers for a given task.

Enables Claude to fetch web pages from the user's own local network connection instead of AI datacenter IPs, including optional authenticated sessions, with verdicts for bot challenges, login walls, and other failures.

Enables a Claude Code agent to fetch remaining 5-hour and weekly subscription usage limits, reset times, and per-window breakdowns as JSON. It provides the same figures as /usage through a local, read-only tool.

Discover and connect your agents with other agents via A2A protocol. Share resources for free or fee.

Aggregates and routes multiple MCP servers with intelligent tool recommendation and batch parallel execution, enabling unified access and efficient tool usage.

A local MCP server that routes QA reviews through deterministic profiles and risk-based escalation while keeping source evidence and final decisions with the host.

This MCP server preserves a Bearer principal on every tool call and SSE frame, rotates vaulted credentials in place, and enforces per-tool allowlists for agents.

mcp-broker is a local Model Context Protocol process broker for MCP clients. Think PgBouncer for MCP: one stable local endpoint in front of many upstream MCP servers. The broker owns upstream startup, reuse, cleanup, profile exposure, status, and safe tool routing. The core idea is simple: do not make every agent session load every upstream tool definition before the user asks a task.

Multi-agent swarm orchestrator for AI workflows. Exposes blackboard read/write, agent dispatch, permission gating (AuthGuardian), token budget enforcement (FederatedBudget), and audit log tools over MCP. Supports 14 AI framework adapters including LangChain, AutoGen, CrewAI, Codex, and A2A — mix any frameworks in one swarm with race-condition-safe shared state.(Traffic light for agents)

* taniwhaai/arai 🦀 🏠 🍎 🪟 🐧 - Policy enforcement for AI coding agents derived from existing instruction files (CLAUDE.md, .cursorrules, .windsurfrules, .github/copilot-instructions.md) — no separate YAML to maintain. Rules with prohibitive predicates (never, forbids, must_not) emit permissionDecision: deny to block tool calls in Claude Code; advisor

Provides blocking await tools to poll commands, URLs, or files until a condition is met, eliminating the need for sleep loops in agent workflows.

Evaluates task suitability for local models before cloud API calls, routing to Ollama or similar to reduce costs.

Open-source permission control plane for AI agents — scan, enforce, and audit every tool call with code-level policies that prompt injection can't bypass.

Production-ready MCP server starter templates in TypeScript and Python. Includes tool, resource, and prompt patterns with Claude Desktop integration configs.

Enables Claude Code to dispatch opencode CLI tasks as background processes with immediate task handle return, status polling, and result retrieval, avoiding tmux and log parsing issues.