MCP servers that ship as code — a repository or a package you install and run locally (usually over stdio). They have no hosted endpoint, so we cannot probe or grade them: instead each one shows its repository, install command, license, stars and when it was last updated.
Collected from GitHub repositories and npm, PyPI, Cargo, Go, Maven, NuGet and RubyGems packages tagged as MCP servers, the official MCP Registry, Docker's MCP catalog, ModelScope, Baidu MCP World, Cline and other public directories; one entry per repository. A server that also offers a hosted endpoint is listed under Remote instead. These entries are not counted in our reachability, verification or readiness statistics.
-
Cross-platform SSH remote execution, passwordless key management, Windows Session 0 desktop launcher, SFTP transfers & MCP server
npm MIT -
Universal user-level MCP workbench for VMs: born in the interactive session, multiplexed panes + session 0/1 shell, spoken to over streamable-HTTP. Windows/Linux/macOS.
-
Terminal and session manager for Windows, Linux and Mac: SSH, SFTP, RDP, VNC, telnet, serial, WinRM, BMC consoles and tunnels in tabs and split panes. Encrypted on disk, with an MCP server and CLI so AI agents run commands without seeing passwords. Rust, Apache-2.0.
Apache-2.0 ★ 15 -
NexaLayerno update data
Managed proxy sessions with hard budget caps, for direct API use and browser automation. Discover current products and prices, create and rotate sessions, monitor health and usage, and enforce owner-controlled spend limits. Early-access launch pricing (subject to change): dynamic proxies $0.23/GB; shared static from $1.05/month; dedicated static $1.43–$5.86/month.
smithery -
Cross-platform SSH remote execution, passwordless key management, Windows Session 0 desktop launcher, SFTP transfers & MCP server
MIT -
tabapigithub.com/Lay4U/tabapiupdated this year
🔍 Record browser sessions → Auto-discover hidden APIs → Export TypeScript SDK + OpenAPI 3.0 + MCP Server. Local-first, privacy-safe.
MIT ★ 1 -
Полка — a shelf for everything your AI agents make: pages, prototypes and projects by link, agent session memory, self-hosted. MCP · AGPL-3.0
AGPL-3.0 ★ 2 -
Local, automatic memory for Claude Code. Remembers past sessions in 0.7s for ~2.5k tokens instead of 211k. Nothing leaves your machine.
pypi MIT -
See what your team did with AI: each session, tool call and dollar traced to a person. Self-hosted, Apache 2.0, REST + MCP.
Apache-2.0 ★ 1 -
Long-term memory for AI coding agents — remembers project rules across sessions at $0 AI cost (MCP server, SQLite/PostgreSQL). 90.26% on Snap LoCoMo.
MIT ★ 83 -
Tetheredno update data
Control Mac power modes, prevent sleep, keep your Mac awake, and manage background apps during AI coding, builds, rendering, and local model inference. Tethered lets your AI assistant discover and apply saved profiles for Low Power, Automatic, or High Power mode where supported; display brightness; fan settings; and keep-awake behavior. Profile settings can deprioritize or suspend selected background apps, including policies triggered by memory pressure. Inspect memory pressure, RAM and swap usage, session status, and task safeguard warnings. Finite sessions expire automatically; unlimited sessions require an explicit end. Power settings describe requested controls, not independently measured hardware state. Requires an Apple Silicon Mac running macOS 15 or later, Tethered 1.2.0+ installed and running under the same macOS account, Pro access, AI task sessions enabled, and permissions required by the selected profile. Runs locally over stdio. Intel Macs and cloud-only clients are unsupported. No connection credentials or saved profiles are included.
smithery -
netcafe-sdkgithub.com/mario03690/netcafe-sdkupdated this year
Zero-dependency toolbox for AI agents: web search, page-to-markdown, diagrams, transcription, PDF, cross-session memory, and one-sentence app deployment, plus a real per-call model cost dataset (CC BY 4.0).
npm MIT -
Universal code & logic calculator for AI models: 51 MCP tools across 31 languages. Rust-sandboxed execution with verdicts, sessions, artifacts, exact arithmetic, and verified translation/optimization. No LLM, no gateway, no telemetry: the caller is the model. Apache-2.0.
Apache-2.0 ★ 5 -
File-bound rules for coding agents. Pin a constraint to the paths it is about; it lives in git, survives session end and tool switches, and re-checks itself. Measured: agents followed pinned rules 18/20 times versus 0/20 without.
npm MIT -
OCDgithub.com/Riko5652/OCDactive
OCD (Omni Coder Dashboard) v5.6.0 — AI memory engine with real semantic embeddings, 25 MCP tools, session guard, trace-to-evidence audit bot, Antigravity intelligence, cost visibility, smart routing, and proactive IDE interception. 100% local, zero API keys.
NOASSERTION ★ 3 -
91.5% fewer tokens on tool output on Headroom's own benchmark, where Headroom removes 61.4%. Separately, the same SWE-bench tasks resolved at 0.71x the cost of plain Claude Code. Plus memory across sessions and subscription quota pacing. Claude Code, Codex, Copilot, Gemini and 12 more. Local, MIT.
npm MIT ★ 581 -
Track what every AI coding agent costs and see the limit coming. Per-session, per-model cost for Claude Code, Codex, Cursor and 37 more tools from your provider's own token counts, plus limit gauges, a dangerous-command guard and dashboard terminals. Free, Apache-2.0, runs on your machine.
NOASSERTION ★ 60 -
AYLENTO · 艾伦兔no update data
# AYLENTO · 艾伦兔 Connect Agents across AI clients using independent identities, private messages, group chats, attachments, follows and block lists. 艾伦兔为不同 AI 客户端中的 Agent 提供独立身份和通信能力,共 36 个 MCP 工具。 Client 0.13.0-beta.4 public beta; MCP runtime 0.13.0-beta.1. 本包为公测版,普通用户注册无需邀请码。 ## Install / 安装 This is a **local stdio MCP bundle**, not a hosted HTTP endpoint. Import the `.mcpb` file into a compatible client. The client must supply **Node.js 24.15.0 or newer**; the bundle includes the JavaScript server and its dependencies, but does not include Node itself. Older bundled Node runtimes cannot run it. No npm installation or API key is required for this bundle. 在支持 MCPB 的客户端导入安装包,并确保它实际使用 Node.js 24.15.0 或更新版本。包内含完整 JS 客户端和依赖,不含 Node 运行环境。安装时为当前客户端选择独立的本地配置名称(默认 `smithery-local`)。 For clients without MCPB import, extract this archive to a permanent directory and configure a local MCP server with command `node`, argument `/absolute/path/to/extracted/bin/aylento.mjs`, and environment `AYLENTO_BASE_URL=https://aylento.com`, `AYLENTO_PROFILE=your-unique-host-profile`. Windows uses an absolute Windows path. Each host's JSON structure differs; see the [official installation guide](https://github.com/popularzb/aylento-plugins/blob/v0.13.0-beta.4/README.md). Do not enable duplicate native-plugin and manual-MCP installations for the same identity. ## Pair and use / 配对与使用 1. Register at [aylento.com](https://aylento.com) and create a separate Agent Chat for this host. 2. Ask: “查看我的艾伦兔连接状态,未连接才开始配对。” The Agent should call `aylento_connection_status` first and `aylento_connect` only when disconnected. 3. Personally approve the returned pairing page and code in your browser. Never paste passwords, session tokens or private pairing links into public listings or issues. 4. Ask the Agent to call `aylento_finish_connection` once after approval. Pending means wait for the person, not continuous polling. 5. Ask “查看我的艾伦兔新消息” or “查看当前身份的黑名单”. To send, specify the recipient's exact rabbit ID and the message; verify the recipient before sending. Use both a separate profile **and a separate Agent Chat** per host. A profile only isolates local storage; it does not create a server identity. Pairing an existing identity elsewhere replaces its prior external session. Keep the same profile during an upgrade to retain local authorization and history. The included `skills/aylento/SKILL.md` contains model-facing instructions. MCPB hosts do not necessarily load skills automatically; follow the host's instructions to load it if supported. Codex, Claude Code and Gemini CLI also have [native installation options](https://github.com/popularzb/aylento-plugins/blob/v0.13.0-beta.4/README.md). ## Data and behavior / 数据与行为 Messages are sent through the AYLENTO service. Incoming message history and attachments are saved to local SQLite before delivery acknowledgment; histories do not automatically synchronize between devices. Tool results may enter the selected host/model's context and logs. Read `PRIVACY.md` for destinations and retention details. Before mutual following or the first reply, only one unsolicited private message is permitted. Once the recipient replies, both parties may continue; blocking and rate limits still apply. `REPLY_REQUIRED` means the message was not sent. Reception is not permission to auto-reply, and receiving messages does not guarantee the host wakes a model. Local reception stops when the process or device stops. If a send has an unknown outcome, inspect messages before attempting again. Disconnecting revokes external authorization but retains local history. Uninstalling the bundle does not erase that history. Do not share actual state directories, credentials, messages or attachments in bug reports. The bundle is checked with isolated MCP startup and a 36-tool inventory. This is not a claim of completed real-model messaging tests in every host or on every operating system. Refer to [host verification scope](https://github.com/popularzb/aylento-plugins/blob/v0.13.0-beta.4/HOSTS.md). ## Copyright and updates / 版权与更新 Copyright (c) 2026 AYLENTO copyright holders. All rights reserved. **AYLENTO Client Use License v1.0** applies; see `LICENSE.txt`. This is not MIT or another open-source license. Personal and organizational internal installation and use are allowed; additional program modification, repackaging, external redistribution, resale and sublicensing require separate written permission. Third-party licenses, statutory rights and rights already granted by official platform terms remain unaffected. Keep `THIRD_PARTY_NOTICES.md` with the bundle. Download updates only from official AYLENTO channels, preserve the profile and local data, and install a newer bundle through your host. Merely pushing Git does not update an installed bundle or publish a Smithery release; the distribution build and authenticated publication must also run. [Website](https://aylento.com) · [Official repository](https://github.com/popularzb/aylento-plugins) · [Release downloads](https://github.com/popularzb/aylento-plugins/releases) · [Public support](https://github.com/popularzb/aylento-plugins/issues)
smithery -
agentedgithub.com/frane/agentedno update data
A text editor for LLMs, not humans. Take ed, the line editor that nobody has voluntarily used since about 1975, and rebuild it for an environment where the typing user is a language model. Short verbs, line addresses, no modes, no TUI. What an editor optimises for changes when the user is the model: round trips per task, tokens per command, an editing buffer with a long memory, and an undo tree that remembers the branches the agent abandoned, because that's often where the interesting work was. ## Features - **Fewer round trips.** Read-before-Edit is unnecessary; on conflict the response carries the new content so you reconcile in one call instead of pre-reading every time. - **Branching undo.** Walked-back work stays addressable instead of being thrown away when you pick a different path. - **Three-way merge.** Concurrent agents get a structured conflict response instead of a silent overwrite. - **Atomic batches.** Multi-file refactors run all-or-nothing instead of leaving half-applied state on failure. - **Cross-file moves and regex replace as primitives.** Operations the built-in tools can't express cleanly become single calls. - **Drift detection.** External edits to an open file are folded into the tree instead of being clobbered by the next write. - **Inline diagnostics.** Type errors and lint findings surface on save, not at the next build many edits later. - **Cross-session memory.** Per-file notes persist between sessions and surface inline on the next open. - **Audit log.** Every operation recorded with actor and timestamp, so two agents in one workspace can't argue about who moved the head. ## Multi-workspace (v0.4.0) One MCP server registration handles every project you mention. Tool calls route to the workspace owning the absolute path you pass; each project gets its own SQLite, its own undo tree, its own LSP daemon — lazily spawned on first touch. Desktop hosts (Claude Desktop, Codex Desktop, Cursor) install the server once globally and use it across projects without per-project config. ## Install `ae` must be installed locally; this MCP bundle registers `ae serve` as a stdio server. ```sh brew tap frane/tap && brew install agented # or curl -sSL https://raw.githubusercontent.com/frane/agented/master/install.sh | sh ``` Pure Go, single static binary, Apache 2.0. Source: https://github.com/frane/agented
smithery -
Contextualno update data
## Contextual — Persistent Memory for AI Coding Agents AI coding agents forget everything the moment your session ends. **Contextual** fixes that by giving any MCP-capable AI client — for developers who use multiple AI coding tools at once — a real, persistent memory of your codebase: not just what the code says, but what changed and why. It's a **local-first MCP server**: your code, your index, and every query stay on your machine. The only network call is a lightweight, infrequent license check — no code, ever, leaves your device. Under the hood, Contextual builds a real understanding of your repository, not a text-similarity guess: - **Structural search** — tree-sitter AST parsing + call-graph resolution, so agents reason about how your code actually connects. - **Bitemporal history** — tracks not just when your code changed, but when the AI learned about it. Ask "did this bug fix happen before or after the agent's understanding of this function?" and get a real answer. - **Git-aware retrieval** — temporal, history-grounded context instead of a stale snapshot. - **Fast, self-contained** — index a 50k-line repository in about 5 minutes, then query it in ~94ms median, from a single ~0.8GB install. This extension connects Claude Desktop (or any MCP-capable client) to a Contextual installation already set up and activated on your machine — it does not provision Contextual from scratch.
smithery -
MCP server for Apple Developer Documentation: API reference, Human Interface Guidelines, App Store Connect Help, sample code and 1,700 WWDC sessions with transcripts and code. A fork of Apple Docs MCP.
npm MIT ★ 26 -
MCP server for Apple Developer Documentation: API reference, Human Interface Guidelines, App Store Connect Help, sample code and 1,700 WWDC sessions with transcripts and code. A fork of Apple Docs MCP.
npm MIT -
Make Instagram usable for AI agents. Meta Graph API v21.0 Twin architecture, MCP server, and Agent plugin for stories, posts, DMs, comments, and anonymous Ghost Mode.
MIT