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.
-
cheap-llm-mcpgithub.com/stblackcat/cheap-llm-mcpupdated this year
Save premium-model tokens by routing simple tasks to cheap OpenAI-compatible LLMs.
npm MIT -
Cheap Money-Saving Self-hosted AI agent workspace with tool calling, MCP, multi-model routing, sandboxed execution, multi-agent workflows, and LLM-authored 3D character animation rendered with Three.js.
MIT ★ 195 -
An MCP server that offloads cheap work from your cloud LLM agent to a local Ollama model.
pypi MIT -
code-aidegithub.com/logicia32/code-aideupdated this year
Offload code reading / review / edits to a cheap LLM via MCP — returns concise results to keep your main agent's context lean.
MIT -
freeloadergithub.com/1977-code/freeloaderupdated this year
Claude Code MCP server + skill that offloads cheap, bulk subtasks to free local LLMs via a self-hosted freellmapi router. Save your Opus/Sonnet tokens.
MIT -
MCP server that lets cheap and open-source LLMs solve optimisation problems exactly. OR-Tools proves the plan optimal and an independent check verifies it. No API key, no LLM inside.
MIT -
Copro: an MCP coprocessor for Claude Code. Delegates side tasks (reading files, summaries, first-pass reviews, extraction) to cheap cloud or local LLMs with budget routing, fallback chains, quota awareness and guardrails.
MIT ★ 1 -
Stdlib-only MCP server that offloads mechanical work from Claude Code to free/cheap LLM APIs (Groq, Gemini, OpenRouter) - ~90% fewer Claude tokens
MIT -
Temporal inference cost arbitrage for AI agents — defer non-urgent LLM work to cheap-energy hours, track savings, and learn user flexibility. MCP-native. Built for AGI House Hackathon.
MIT -
websearch-mcpgithub.com/jason920612/websearch-mcpupdated this year
Search-engine-free semantic web search as an MCP server: a cheap LLM navigates the live web, a local embedding model indexes what it walks past. 不依賴搜尋引擎的自建語意網路搜尋 MCP。
MIT -
让 Claude Code / Cursor 多一个便宜的第二意见:一句话调用 DeepSeek、Kimi、GLM、通义千问。支付宝充值即可,无需海外信用卡与 VPN。 | MCP server that gives your coding agent cheap Chinese LLMs as a second opinion.
MIT ★ 1 -
local-llm-mcpgithub.com/HenryLinyy/local-llm-mcpupdated this year
MCP server that lets Claude Code & Codex delegate bounded, token-heavy generation to local or cheap-cloud LLMs — LM Studio · Ollama · vLLM · llama.cpp · DS4, plus DeepSeek/OpenRouter/Groq/Cerebras/MiniMax/Agnes. Keep the frontier agent's judgment, cut the token bill, with RAM guards and a text-only safety boundary.
MIT ★ 3 -
Mult-Hands-Eyes-MCPgithub.com/MicroHEROX/Mult-Hands-Eyes-MCPupdated this year
MCP server that gives online LLMs local hands and eyes: OpenAI-compatible local inference (KoboldCpp / Unsloth / llama.cpp / LM Studio / Ollama) for cheap text work and vision work (OCR / image analysis / comparison).
npm MIT -
Cheap eyes for an expensive agent: an MCP server that delegates bulk reading to cheap OpenRouter models and mechanically checks every quote and line reference against the source.
npm MIT ★ 4 -
50x your usage with cheap-ag. Opus/Astra plans, a dirt-cheap model writes every line, and tests decide what ships. Your $20 plan stops hitting limits. One MCP server for Claude Code & Codex.
npm MIT ★ 332 -
glyph-mcpgithub.com/k1y0miiii/glyph-mcpupdated this year
Token-cheap web search and fetch tools for Claude, backed by glyph
MIT -
sprout-mcpgithub.com/mepsopti/sprout-mcpupdated this year
Model-tiered content pipeline MCP server — cheap models seed, expensive models verify
pypi MIT -
frugalgithub.com/SRKRZ23/frugalupdated this year
Run AI agents cheap, local, and verified — metering, cost routing, cache, eval, MCP cost server. Offline, zero-dep, Apache-2.0.
NOASSERTION ★ 1 -
ltpmcpgithub.com/gashel01/ltpmcpupdated this year
Compile-once planning (LTP) — turn a goal into a verifiable, token-cheap plan with static safety verification before execution. Python lib · CLI · MCP server · FastAPI.
Apache-2.0 ★ 1 -
Cheap, confidence-scored question answering over knowledge graphs: each hop is a calibrated classification decision. Library, CLI, and MCP server.
pypi Apache-2.0 -
A master orchestrator delegates routine coding to cheap apprentice models — local Qwen + Gemini — behind one MCP server, then auto-verifies, corrects, and learns from the results.
MIT ★ 1 -
deepseek-delegate-mcpgithub.com/trixmix821/deepseek-delegate-mcpupdated this year
Local MCP server that lets Codex (GPT Sol) delegate boilerplate and bulk code generation to DeepSeek V4 — hybrid planner + cheap executor.
MIT ★ 2 -
OpenRouter-powered read-only subagent MCP for cheap code search, file reading, and project investigation.
npm MIT -
MCP server that lets Claude Code hand bounded coding tasks to cheap models, behind tests, a second-model review and cost caps.
-
One lead model plans and reviews, cheap local or remote workers write the code in parallel. Plugin for Claude Code, MCP server for Codex, extension for Pi. Zero dependencies.
MIT -
MCP server for azair.eu — cheap flight search across European low-cost airlines with flexible dates, multi-airport, and metro-area search.
★ 3 -
A tiny MCP stdio server that exposes DeepSeek as a cheap supervised worker for Claude Code, Codex, or any MCP client.
MIT ★ 56 -
An Obsidian-style vault your AI agents can read — and an orchestrator that lets a cheap model write while your expensive model only reviews.
MIT ★ 2 -
Fast, cheap, probability-scored decisions for coding agents, scripts and CI. CLI + MCP server + Claude Code plugin for the OpenAI Decisions API and Jev (TypeSafe AI).
npm MIT ★ 8 -
An MCP server that delegates mechanical code edits to a cheap worker model, validates them through a chain of gates, and applies them atomically. Supports PHP & Python.
Apache-2.0 -
cheaplanegithub.com/millennialdreamer/cheaplaneupdated this year
Offload replaceable grunt work from your premium AI agent to cheap models via MCP. Keep your subscription on the main thread; billing is physically isolated. Auto-routing + savings ledger + one-command setup. MIT.
pypi MIT ★ 1 -
AI assistant and coding agent that picks the right model for every task: fast, cheap models for routine work, frontier models for the hard parts. Plans, edits and checks code as reviewable diffs. Desktop app, CLI and MCP server.
NOASSERTION ★ 2 -
mailctxgithub.com/hannokuegler/mailctxupdated this year
📬 Give your AI agent your inbox. Provider-agnostic IMAP → clean, token-cheap Markdown (60–98% fewer tokens), plus an MCP server so Claude Desktop can search, read and draft your mail. Reads and drafts — never sends.
MIT -
Claude plans and reviews, cheap models type. An MCP server that hands coding jobs to OpenCode workers in isolated git worktrees — hard price ceilings, a daily budget, and nothing merges until you have seen the diff.
MIT -
把订阅的模型、按量付费模型和生图生视频模型统一管起来:Claude Code、Codex、WorkBuddy、Hermes 等智能体在哪都能用,按难易自动挑模型,失败自动换下一个,能省就省。A local model dispatcher for AI agents: subscription first, cheap models for chores, strong models for hard problems.
MIT ★ 1 -
Repo-aware directive generator for AI coding agents: scans a repository, a cheap model assesses its state, and a capable model writes the directive. Python CLI with an optional MCP server (stdio or SSE) and OpenRouter or local-router backends.
pypi -
Workspace-pinned local code KB + MCP bridge giving AI coding agents AST-precise, ranked, token-cheap access to your codebase — ~99.9% less context than grep-and-read. Local, offline, 12 languages.
NOASSERTION ★ 2 -
glm-mcp-claudegithub.com/djerok/glm_mcp_claudeupdated this year
GLM (Zhipu/Z.ai) as a cheap, full-capability subagent for Claude Code — auto-routing between Opus and GLM, a file-editing agent with diff/dry-run/git-revert oversight, and a one-command installer.
npm MIT -
cerebro-code-memorygithub.com/marcodavidd020/cerebro-code-memoryupdated this year
Persistent code-knowledge memory for AI chats. An MCP server that caches a codebase's structure and summaries in a local SQLite "brain" so sessions query it instead of re-reading files — token-cheap, 100% local.
pypi MIT ★ 6 -
Suber-Agent-Teamgithub.com/stephenpham68/Suber-Agent-Teamupdated this year
MCP server that offloads Claude Code sub-agents to a cheap model (Haiku / any OpenAI- or Anthropic-compatible API) - run 20-100 parallel agents for fan-out, map-reduce & research without draining your Opus / Claude Max quota.
MIT ★ 3 -
A strong model (Claude or ChatGPT, on your subscription) leads as the lead engineer; cheap models (DeepSeek, GLM, Kimi, MiniMax, Qwen) write the code in isolated git worktrees. Budgets, live dashboard, MCP server, project memory. Desktop app and web service.
MIT -
glm-mcpgithub.com/djerok/glm-mcpupdated this year
GLM (Zhipu/Z.ai) as a cheap, full-capability subagent for the Claude Code app — works on a subscription Claude (no API key for the main agent), auto-routes between Opus and GLM, file-editing agent with diff/dry-run/git-revert, one-command npx install.
npm MIT ★ 8 -
Stop your coding agent burning context on file reads. Delegates big reads to a cheap worker model and returns only the answer — 87-99% saved, measured per call. A hook enforces it in Claude Code; the MCP tools work in any client.
MIT -
Zero-infrastructure, local-first memory for AI agents. One SQLite file — no Postgres, Neo4j, or cloud. Bi-temporal supersession (updates, never deletes), four-channel hybrid retrieval, and a drop-in MCP server. Runs offline on cheap models. Apache-2.0. Bring your own benchmark.
Apache-2.0 ★ 14 -
MCP server that lets any agent delegate to headless coding-agent CLIs — Codex, Grok, Claude Code, OpenCode, Kimi, Muse (Cursor opt-in): cheap/fast delegation, codebase exploration, surgical reads, filtered command output and web lookup, with model + effort control, a mandatory bwrap sandbox and session follow-up.
MIT ★ 4