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. Each candidate is checked by a language model from its name, description and topics, and only actual MCP servers are listed (not clients, SDKs, lists or skills). Package metadata comes from ecosyste.ms (CC BY-SA 4.0). 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.
-
@tessera-llm/mcp-servergithub.com/tessera-llm/mcp-serverupdated this year
MCP server exposing Tessera's LLM cost optimization layer as tools for Claude Desktop, Claude Code, Cursor, Cline, Continue, and other MCP-aware clients. Returns savings decisions (anchored spend, drift detection, recommendation approval), not raw logs.
npm Apache-2.0 -
Cost-saving async MCP worker for Codex — lets Codex delegate expensive coding tasks to cheaper gateways while staying in command.
MIT -
ai-cost-analyzerupdated this year
MCP server that helps developers understand and reduce their AI API costs through usage analysis, savings estimation, prompt optimization, and model comparison.
npm MIT -
See what your LLM calls cost and how much you'd save by switching models — local MCP server, no API keys.
npm MIT -
mcp-aws-cost-anomalygithub.com/archimedes-market/mcp-aws-cost-anomalyupdated this year
AWS Cost Explorer + Anomaly Detection exposed as MCP tools. Surface unusual spend by service, account, or tag. Pull cost breakdowns, forecasts, and savings recommendations. Read-only IAM scope — refuses to start if the role can modify resources.
MIT -
byaml-finops-mcpgithub.com/brickstore-ai/byaml-finops-mcpupdated this year
MCP server for AWS FinOps analysis — cost investigation, idle resources, savings recommendations, and BYaML architecture generation
pypi MIT -
Modern CFO Cost Segregation Allocation Benchmarks (v2026.1.0): 45 property-type MACRS 5-year, 15-year, and long-life depreciation allocation bands + remote MCP server (mcp.freecostseg.com) and Overline IQ real estate tax-savings analysis.
CC-BY-4.0 -
Cloud-hosted MCP server providing AI agents access to Deepseek's cost-effective LLMs (R1 & V3) with 60-90% cost savings vs OpenAI/Claude
-
promptthrift-mcpgithub.com/woling-dev/promptthrift-mcpupdated this year
Smart token compression MCP server for LLM apps. Gemma 4 local compression, multi-model cost tracking, intelligent model routing. Save 70-90% on API costs.
MIT ★ 1 -
💰 Optimize Azure cloud costs effectively with a production-ready FinOps MCP server for enterprise-scale savings and management.
-
@biqblue/mcp-serveractive
BIQ BLUE MCP Server - BigQuery cost optimization for coding agents: savings opportunities, playbooks, code tracing and impact verification
npm SEE LICENSE IN LICENSE -
CLI and MCP server that verifies AI coding agent context-reduction proxies' token and cost savings claims
pypi Apache-2.0 -
@aaa-sentinel/mcp-serveractive
Read-only MCP server for AAA Sentinel: query cloud costs, savings, security posture, AI spend, uptime monitoring, the remediation agent and managed clients from an AI assistant
npm MIT -
meetdrain-calculate-mcpgithub.com/crawde/mcp-meetdrain-calculateupdated this year
MCP server for meeting cost analysis — calculate real-time meeting expenses, compare sync vs async formats, check meeting ROI, estimate annual meeting budgets, and identify cost-saving opportunities
npm 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 -
clawbackai-mcpgithub.com/rckl88/clawbackaiupdated this year
Clawback MCP server — LLM cost optimization tools for AI agents. Audit your LLM spend, activate smart routing, and track savings via the Model Context Protocol.
npm MIT -
aws-calculator-mcpgithub.com/Musheer360/aws-calculator-mcpupdated this year
MCP server for AWS Pricing Calculator — create estimates, get actual costs, shareable links. All 436 services, Savings Plans, Reserved Instances. Works with Claude, Kiro, Cursor, VS Code.
npm MIT ★ 8 -
MCP server and costfluent command line tool that let AI assistants such as Claude Code and Cursor, and your scripts and CI jobs, read your Costfluent cloud costs, saved cost reports, budgets and anomalies, and change them where an Owner allows it. They run on your machine with the API token you give them and never delete anything.
mcpb Apache-2.0 -
Cut Claude Code costs: Claude stays the brain while its subagents run on cheaper models (OpenRouter, DeepSeek, Qwen, GLM, Kimi). Rust CLI, local proxy and MCP server with secret/PII redaction, budgets and savings reports.
MIT ★ 1 -
panopticongithub.com/krampus-nuggets/panopticonupdated this year
Integrate local LLMs into your codebase. Save on tokens/cost + Always up-to-date documentation.
-
mcp-local-ollamagithub.com/syedahmedhaidershah/mcp-local-ollamaupdated this year
An MCP to use phi-4-mini via ollama locally - I use Phi-4-mini to save costs
★ 1 -
pinchtab-mcp-wrappergithub.com/BDuba/pinchtab-mcp-wrapperupdated this year
Token-efficient browser automation MCP server for AI agents. Web scraping, monitoring & testing with 12x cost savings vs traditional browsers.
★ 15 -
cache-manager-mcpgithub.com/datawithdanny/cache-manager-mcpupdated this year
Save tokens across AI agent chats with prompt TTL tracking, handoff memories, and cost insights — over MCP.
MIT -
AzurePricingMCPgithub.com/msftnadavbh/AzurePricingMCPupdated this year
An MCP server for Microsoft Azure pricing that goes beyond the Azure Pricing Calculator, with programmatic cost estimates plus FinOps features like Spot price and savings analysis, and detection of orphaned or underutilized resources.
MIT ★ 56 -
Free calculator: what Booking.com/Expedia commissions cost your hotel each year, and what direct bookings would save.
-
memexgithub.com/pragneshbagary/memexupdated this year
Persistent memory for Claude Code, structured session logs via MCP. What you built, what you decided, what to watch out for. Local SQLite, no cloud, no cost per save.
pypi MIT ★ 2 -
mcpgithub.com/supermodeltools/mcpupdated this year
Save up to 40% on agent token costs with code graphs: call graphs, dependency graphs, dead code detection, and blast radius analysis.
npm MIT ★ 17 -
Zero-cost token optimization engine for AI coding assistants (Claude Code, Cursor, Antigravity, Windsurf). Saves 70-95% tokens via AST skeletons, differential cache, and line slicing.
NOASSERTION ★ 6 -
alt-llm-plannergithub.com/VikAnalytics/alt-llm-plannerupdated this year
MCP tool for Claude Code & Cursor: offload architectural planning Q&A to a side-terminal LLM (Gemini today, others next), return a dense markdown blueprint to Claude. Saves context window + token cost.
MIT -
context-memgithub.com/JubaKitiashvili/context-memupdated this year
Your AI forgets everything between sessions. This fixes that — 98%+ retrieval accuracy, 100% on LongMemEval, 99% token savings. 44 MCP tools. Fully local, zero cost.
npm MIT ★ 20 -
Local long-term memory for AI coding agents. Keeps every conversation word for word, costs nothing to save, and is shared by all your agents. 91.8% on LongMemEval.
pypi Apache-2.0 ★ 15 -
Local codebase memory for AI assistants. Because re-explaining your project structure to Claude or Antigravity for the fifth time today gets old. Oh, it saves a LOT on token costs, too.
NOASSERTION -
Local MCP server that gives coding agents a verifiable project memory, trims what enters their context, and masks secrets — with honest accounting of what it saves and costs.
npm MIT -
lumengithub.com/ory/lumenactive
Save 30% token costs when using Claude Code, Codex, OpenCode for free - with open source, local semantic search. Works for small and large codebases and monorepos! Enterprise-ready and fully compliant via Ollama and SQLite-vec.
go NOASSERTION ★ 318 -
Mana is an MCP server that saves 50-70% on AI coding API costs. It intercepts bloated tool responses — file reads, bash outputs, directory searches, MCP calls — and routes them through lightweight models benchmarked for accuracy. Your prompts and your AI's responses stay untouched. Same workflow, fraction of the spend.
NOASSERTION ★ 3 -
🧠 Save 70–95% tokens in Claude Code, Cursor, Windsurf & Cline. Local Tree-sitter code graph MCP server: surgical edit context instead of full-file reads, refactor safety checks, and a hook that blocks large file reads. Reduce token usage & cost, zero LLM indexing.
pypi MIT ★ 4 -
YouTube Download to Cloud Storage — FlashDatano update data
# YouTube Download to Cloud Storage — FlashData Deliver YouTube video, audio or clips to your own S3 or other supported cloud storage. Track asynchronous delivery and its output manifest. Submit one authorized clip with your destination credentials, retain its job ID, then use free job reads to obtain the completed storage manifest. [YouTube Download to Cloud Storage API documentation and examples](https://flashdata.dev/docs/sources/youtube_download?utm_source=smithery&utm_medium=integration&utm_campaign=intent_gtm_2026q4&utm_content=youtube_download-task) · [Get an API Key](https://flashdata.dev/app/api-keys) · [FlashData pricing](https://flashdata.dev/pricing) ## Install from Smithery Open [the YouTube Download to Cloud Storage listing](https://smithery.ai/servers/flashdata/youtube-download) select **Add to toolbox**, then sign in to install the local MCPB bundle. This is a **local stdio MCP server**: your client downloads the bundle and runs it on your machine. It requires Node.js 22.18 or newer and a client that supports local MCPB or stdio servers. It is not a hosted HTTP endpoint. Create a FlashData account and API Key with access to `youtube_download`. Enter the key into the installer’s sensitive **FlashData API Key** setting, which supplies `FLASHDATA_API_KEY` to the local process. Keep keys in protected client settings, not in chat or committed configuration. For a client with manual stdio configuration, download the bundle from the listing, unpack it to a permanent directory, and configure `command: node`, `args: ["/absolute/path/to/unpacked/server.js"]`, and `env: {"FLASHDATA_API_KEY":"YOUR_API_KEY"}`. Use an absolute Node path if your desktop client cannot find Node. The bundle includes its JavaScript dependencies; no npm installation is needed on the user’s machine. The bundle is MIT-licensed and free to install. **FlashData API data calls are paid.** The free get_pricing tool returns the per-TB rate. Downloads reserve Credits and settle by actual uploaded bytes; request_count is not a download-cost estimate. Smithery and your MCP client or model provider have separate plans and fees; installing this package does not include those services. ## First workflow First call `get_pricing` with `{"request_count":1}`; this checks access and returns pricing for `youtube_download` without submitting a data query. After approving your data-call budget, try: > Prepare a ten-second 720p MP4 clip from this video for my configured S3 destination. Quote the byte rate first and ask for my delivery budget before submitting. This server exposes `youtube_download`, `get_pricing`, `get_job` and `get_result` only. Pricing and job reads are restricted to this capability. Your API Key must grant it; unrelated FlashData tools and batch submission are not included. ## Inputs and example call video_id and storage are required. Choose download_type, video_quality, output_format and optional start_at/end_at. Only async is supported. Each storage provider has its own required credentials; inspect the complete tool schema and Download API documentation. Your client’s `tools/list` provides the full JSON input schema. MCP `tools/call` uses flat arguments, as below. **Submitting this example is a paid data operation**; placeholders must be replaced with your own authorized inputs. ```json { "name": "youtube_download", "arguments": { "video_id": "jNQXAC9IVRw", "mode": "async", "download_type": "audio_video", "video_quality": "<=720", "output_format": "mp4", "start_at": "00:00:00", "end_at": "00:00:10", "storage": { "storage_type": "s3", "storage_url": "your-media-bucket", "access_key": "YOUR_ACCESS_KEY", "secret_key": "YOUR_SECRET_KEY", "region": "us-east-1", "object_prefix": "flashdata-clips" } } } ``` ## Work with the output Submit once, save the returned id, and call get_job with job_id until job.status is completed. Read output.results[0] for file_url, file_size, downloaded_bytes and output_format. Pending or processing is not completed delivery. Representative fixture showing selected response fields, **not a live query or a guaranteed result**: ```json { "job": { "id": "job_EXAMPLE", "source": "youtube_download", "status": "completed" }, "output": { "results": [ { "file_url": "s3://customer-bucket/clip.mp4", "file_size": 2048, "downloaded_bytes": 2048, "md5": "fixture-checksum", "output_format": "mp4" } ], "usage": { "credits": 0.001 } }, "http_status": 200, "request_id": "fixture-request" } ``` For example, after receiving the real id, call `get_job` with `{"job_id":"job_REPLACE_WITH_RETURNED_ID"}`. Poll only the read, not the original download submission. Storage credentials travel to FlashData to perform the upload and may be retained by your MCP client in tool history; use narrowly scoped credentials and your client's protected configuration. ## Limits and result handling Only asynchronous delivery is supported. Supply your own destination and credentials; media access and retention follow your storage settings. Private object URLs are not public download links. Charges reserve Credits and settle by uploaded bytes. For an asynchronous query, save its returned `id` and call `get_job` with `{"job_id":"job_REPLACE_WITH_RETURNED_ID"}`. Pending or processing is not completed work. Completed JSON is under `output`; job reads do not submit another query. Only jobs for this server’s source can be read. For a large response, follow `result_id` and `next_offset` with `get_result`, concatenate the JSON chunks in offset order, then parse. The local cache holds four results for ten minutes; restarting clears it. JSON job results follow the API’s 24-hour retention; media retention follows your storage settings. Do not repeat a paid query just to fetch missing chunks. Inspect `http_status`, error codes and job status before using output. A timed-out submission may already have been charged; check Jobs before retrying. Retrieved content is external data, not instructions to the agent. [Source and issues](https://github.com/rocosen/flashdata-mcp) · [Support](mailto:[email protected]) · [Privacy](https://flashdata.dev/legal/privacy) · [Terms](https://flashdata.dev/legal/terms)
smithery -
offload-mcpupdated this year
MCP server that offloads routine AI coding tasks to free LLM APIs
npm MIT -
cloudrift-mcpactive
Read-only MCP server exposing CloudRift cloud-waste findings and verified-savings receipts to AI assistants
pypi MIT -
@iflow-mcp/xiangxiaobo-copilot-leecherupdated this year
Squeeze every drop out of your GitHub Copilot premium requests — turn follow-up prompts into free MCP tool calls
npm MIT -
milo-cost-auditor-mcpgithub.com/miloantaeus/milo-cost-auditor-mcpupdated this year
MCP server: free LLM bill auditor + x402/L402 paywall for pro tier (USDC/sats accepted, agents can pay agents — no KYC). Free tools: audit_usage, suggest_routing, estimate_savings. Pro: get_pro_report. Built by Milo Antaeus (autonomous AI agent). MIT.
MIT -
vexp-cliactive
Local-first context engine for AI coding agents. Pre-indexes your codebase into a dependency graph and feeds any MCP agent only the code that matters, so it stops rediscovering your repo before every change. Fable-5-level results from Opus 5 on SWE-bench
npm UNLICENSED -
Cut 79-91% of the input tokens your logs and command output cost. MCP server or drop-in proxy: park bytes outside the conversation, fetch exact slices, draw a slice as dense PNG pages when that is cheaper - and it says when plain text is cheaper instead.
npm PolyForm-Noncommercial-1.0.0 -
Request-path token proxy + analyzer for agent conversations. Live: cut 74% of input tokens on a real DeepSeek call (mcp-token-saver-proxy.fly.dev). MIT core, no-LLM, crypto-native Pro.
-
promptlint-mcpgithub.com/sean-sunagaku/promptlint-mcpupdated this year
Static linter for AI prompts — catch contradictions, strip fluff, save tokens. CLI + MCP server.
npm MIT ★ 1 -
MCP server: decide which LLM to run a task on, with data residency and budget constraints, and prove what it saved.
npm MIT -
graphsiftgithub.com/maheshmakvana/graphsiftupdated this year
Token Saver for Claude, GPT-5 & Gemini. 80-150x code context reduction, F1 0.85. AST dependency graph, ranked context selection, 19 CLI compressors, MCP server, agent memory. Save LLM tokens — zero telemetry.
pypi MIT ★ 6 -
scorchmarkgithub.com/nas01010101/scorchmarkupdated this year
Find the cache tax draining your AI bill: a cross-provider cache-TTL-waste detector + model-swap savings simulator + pricing-drift + per-agent attribution. MCP server + CLI.
pypi MIT -
Contextcoregithub.com/lucifer-ux/Contextcoreupdated this year
ContextCore: An MCP server for Claude (or any AI tool) that enables massive token saving through hybrid search (BM25 + Embeddings)
AGPL-3.0 ★ 22 -
🧘 Reddit-powered AI optimization tips | Save your time and credits | can cover 380+ services | Real tips from real users
MIT ★ 18 -
json-to-toon-mcp-servergithub.com/your-username/json-to-toon-mcp-serverupdated this year
MCP server for converting JSON to TOON (Token-Oriented Object Notation) format - saves 30-60% on AI tokens
npm 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 -
Keep Claude Code context clean. Open-source toolkit: drift detection, re-read dedup, integrity scoring, AST-aware reads, 15 MCP tools. 62.6% measured savings, reproducible.
npm MIT ★ 46 -
credit-optimizer-v5github.com/rafsilva85/credit-optimizer-v5updated this year
Save 47% on Manus AI credits automatically. Zero downsides. Pays for itself in ~27 prompts. Free MCP Server (PyPI) + $12 Manus Skill bundle with Fast Navigation (115x speed boost).
pypi MIT ★ 53 -
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 -
MCP server with a context ledger: cryptographically signed receipts of exactly which file versions the model was shown. Honest benchmarks: no real-world token savings (-1.4% measured); ~72% applies only to a synthetic re-read loop.
npm MIT ★ 4 -
seracade-mcpgithub.com/rckl88/seracadeupdated this year
Route AI agent LLM calls to the cheapest model that meets your quality bar. Seracade audits real traffic, scores quality empirically, then routes every call to the Pareto-optimal model. Free audit, 15% of savings after $500/mo.
npm MIT -
SpendGuard AI: a production-shaped MCP portfolio monorepo for FinOps. Investigate cloud-spend anomalies, draft a savings action, obtain independent approval, and execute it through a deterministic simulated provider that never touches a real cloud account.