Ai And Ml for AI agents
100 security-reviewed ai and ml listings, skills, MCP servers, and toolkits you can install into Claude Code, Cursor, and other agents with one command.
Td Feature Survey
Survey one surface of tdmcp (artist controls, library/packaging, CLI/DX, AI/LLM, or TouchDesigner depth) for candidate NEW features — inventory what exists, cross-check the roadmap, apply the gap-finding lenses, vet each idea, and emit a structured, novelty- and confidence-labelled candidate list to _workspace/discovery/. Use when a td-surveyor agent is scouting a surface during the feature-disco…
Connectors Directory Spec
Authoritative reference for submitting tdmcp to the Anthropic Connectors Directory as a Claude Desktop Extension (MCPB). Use whenever drafting, checking, or fixing the submission — it defines the two submission paths, why tdmcp is a Desktop Extension (never a remote connector), every form field, the rejection-causing gates, and the local-only compliance story. Read this before writing any submiss…
Tdmcp Feature Discovery
Survey the whole tdmcp project and produce a prioritized list of NEW features it could implement — across artist controls, CLI/DX, AI/LLM integration, and TouchDesigner depth. Use whenever the user wants to brainstorm, discover, survey, list, or audit what features/tools/effects/controls/commands/prompts/capabilities tdmcp *could* add, asks 'what could we build / what's missing / what are the opp…
Tdmcp Submission
Orchestrates the tdmcp Anthropic Connectors Directory (Desktop Extension / MCPB) submission prep — designs the package, writes the privacy page, migrates .dxt→.mcpb, drafts all form answers, and QAs the result. Use for ANY work on the tdmcp directory/marketplace submission, including re-runs after a rejection ("update the submission", "redo the form answers", "re-migrate the bundle", "fix the pri…
Api Scam Hunter
>
Axure Rp Html Ingest
MUST USE when the task mentions Axure, RP, published HTML prototype, html_export, index.html prototype packages, data/document.js, files/*/data.js, extracting prototype structure, or converting prototype pages into LLM-ready context. Ingest and parse Axure RP exported HTML prototypes into structured JSON and Markdown.
Scanpy Sc Analyzer
Autonomous single-cell RNA-seq quality control filtering, Harmony batch-effect correction, Leiden clustering, UMAP visualization, and marker gene annotation.
Alphafold Pocket Evaluator
Parses AlphaFold2 PDB files, computes per-residue pLDDT confidence scores, and evaluates Solvent Accessible Surface Area (SASA) of active site pockets.
Pubmed Rag Synthesizer
Vectorizes PubMed abstracts using PubMedBERT, builds FAISS indices, and instructs agents to synthesize clinical evidence with exact PMID citations.
Diffdock Virtual Screener
Runs DiffDock generative diffusion models for blind protein-ligand docking against AlphaFold structures and ranks candidates by confidence scores.
Pydeseq2 Bulk Rna
Automated negative binomial differential gene expression analysis, log2 fold-change calculation, p-value adjustment (FDR), and Volcano plot generation.
Jarvis Ai
Local voice-driven AI operator (JARVIS) for Linux & Windows — Claude Pro / OpenRouter / Ollama, wake word, piper TTS, Iron Man-style HUD, Hyprland integration, WhatsApp & Telegram, Agentic Task Engine with live progress tracking.
Gatk4 Somatic Variant Caller
Executes GATK4 Mutect2 tumor-normal paired variant calling, applies LearnReadOrientationModel, and runs FilterMutectCalls for somatic SNV filtering.
Ensembl Vep Variant Annotator
Parses multi-sample VCF files, queries Ensembl VEP REST API, filters ClinVar pathogenicity ratings, and maps driver mutations to FDA targeted drugs.
Qiime2 16s Microbiome
DADA2 amplicon sequence variant (ASV) dereplication, Alpha/Beta diversity index calculation, SILVA taxonomy classification, and 3D PCoA projections.
Rdkit Qsar Pharmacophore
Computes 2048-bit ECFP4 Morgan fingerprints from SMILES, trains LightGBM regressors for pIC50 prediction, and extracts SHAP feature attributions.
Tcga Cibersort Deconv
Downloads TCGA RNA-Seq datasets via TCGAbiolinks, normalizes log2(TPM+1) counts, and calculates tumor-infiltrating immune cell fractions (CIBERSORT).
Bazi
Use when the user asks for bazi, 八字, Four Pillars, BaZi, Chinese astrology birth chart, 生辰八字, 命盘, day master, or interpretation from a Gregorian birth date and time.
Xiaoliuren
Use when the user asks for Xiao Liu Ren, 小六壬, xiaoliuren, quick symbolic timing reflection, daily guidance, folk-style yes/no reflection, or interpretation from lunar month/day/hour inputs.
Tarot
Use when the user asks for tarot, a card pull, a card reading, symbolic reframing, decision reflection, creative blocks, project reflection, or interpretation of a user-provided tarot draw.
Iching
Use when the user asks for I Ching, Yi Jing, Zhouyi, 易经, 周易, hexagrams, coin casting, yarrow-style casting, change analysis, or interpretation of user-provided line values.
Better Prompt
Optimize and refine AI prompts through a two-phase expert pipeline. Phase 1 (Lyra) builds a structured draft from rough ideas, Phase 2 (Meta) deeply refines it with advanced prompt engineering techniques. Use this skill when the user wants to improve a prompt, write a better system prompt, optimize LLM instructions, or craft high-quality AI prompts. Also trigger when users ask about prompt engine…
Resource2skill
把人类创作的资源（教程视频/文章/代码仓库/成品参考）蒸馏成可复用的 skill——Resource2Skill 论文 (arXiv:2606.29538) 构建算子的本地实现。触发场景：用户说"把这个视频/教程/文章变成 skill"、"蒸馏这个资源"、"入库这个教程"、给出教程链接并希望沉淀为规范/技能，或要为某领域批量建技能库时。产物形态按目标域适配（Claude Code skill / PPT 版式技能 / n8n 工作流模式等）。
Init Merlean
Set up the MerLEAN system in this repo (replaces the old src/deploy.sh) — build the shared Python .venv from src/requirements.txt, write the root .mcp.json for the lean-lsp MCP server, ensure an OpenAI key in .env, and prepare the root Lean + Mathlib workspace (lake update / cache get / build). Use when asked to install, set up, initialize, or (re)deploy MerLEAN, or when a fresh clone needs its r…
Plan Graph
Read, edit, and SEMANTICALLY SEARCH the formalization statement plan stored in the Mem0-g plan graph (the replacement for statements.json). Use whenever you need to inspect, query, or modify statements, dependencies, status, or informal notes; run the Lean dependency sync; or find statements/notes by MEANING (LeanSearch-v2-style two-stage retrieval: OpenAI embedding recall → LLM rerank) — e.g. to…
Skim
Runtime token proxy + intelligence dashboard for LLM tools. Intercepts API calls, strips waste, tracks costs — for individual developers and Fortune 500 teams.
Turbovec MCP Tiar
Turbovec MCP Server: A local Retrieval-Augmented Generation (RAG) tool that allows AI to seamlessly read, chunk, remember, and semantically search through large project codebases.
Godot Plugin
Control Godot Editor via OpenClaw Godot Plugin. Use for Godot game development tasks including scene management, node manipulation, input simulation, debugging, and editor control. Can modify, save, and delete scenes/nodes and capture project metadata and screenshots — destructive operations should be confirmed with the user. Use only on explicit Godot Editor requests like inspecting scenes, crea…
Person To Skill
Builds or updates a source-grounded advisor skill from one person's books, documents, local transcript files, and YouTube video URLs. Extracts PDF/EPUB/DOCX/HTML/Markdown/text/RTF/MOBI/AZW sources, retrieves available YouTube captions, reconciles evolving or conflicting ideas, and generates a compact `/ask-name`-style skill with citations, framework notes, decision rules, source provenance, and c…
Docsift
Convert documents once. Give agents only what they need — PDFs and Office files to clean Markdown, chunks and local search.
Trapstreet Task Scaffold
Design and scaffold a NEW trapstreet.run task to evaluate a given agent/skill/tool -- the reverse of trapstreet-solution-scaffold (which builds a solution against an existing task). Generates the mechanical parts (traptask.yaml, judge.py/grader.py matching the TRAPTASK_MANIFEST contract, build_cases.py's validate-then-render pipeline) and guides the judgment-heavy parts through a structured inter…
Trapstreet Solution Scaffold
>-
Unreal Plugin
Control Unreal Engine Editor via OpenClaw Unreal Plugin. Use for Unreal development tasks including level/actor management, transforms, PIE control, debugging, input simulation, and console commands. Can create/delete actors, save levels, run console commands, simulate input, and capture viewport screenshots and logs — destructive operations should be confirmed with the user. Use only on explicit…
Life Paths
Structured process for mapping realistic long-term life and career paths for a person, grounded in their actual record and finances rather than generic advice. Use this whenever someone asks for help figuring out what to do with their life or career, wants a long-term or phased plan, mentions retirement or FIRE timing alongside career choices, asks whether they could realistically become or achie…
Financial Planning
Build a grounded, verified long-horizon financial plan for a person - accumulation, financial-independence timing, drawdown, and the decision rules to run it by. Use this whenever someone asks when they can retire or go work-optional, wants a FIRE or retirement plan built or reviewed, brings a spreadsheet or planner document to audit, asks about safe withdrawal rates, savings schedules, drawdown…
Writing Axes
Audience-and-goal-first drafting, review, and rewriting for any human-facing prose. Use whenever the user asks to write, rewrite, draft, critique, review, or edit a document, including blog posts, READMEs, papers, emails, docs, runbooks, error messages, release notes, incident reports, announcements, and forum posts. Also use when asked for a title, headline, or subject line, or to check somethin…
Huggingface
>
Huggingface
>
Huggingface
>
Cowork Tasks
The kanban that fills itself in. Watches your email, Slack, meetings, Linear, Jira - reads, writes, and moves cards as work evolves. Open-source, MIT, local-first. Built on Claude Cowork's Live Artifacts.
Unity Plugin
Control Unity Editor via OpenClaw Unity Plugin. Use for Unity game development tasks including scene management, GameObject/Component manipulation, debugging, input simulation, and Play mode control — including arbitrary C# execution (script.execute) and reflection-based editor calls, which can modify scenes, assets, and settings. Use only in trusted local projects; destructive operations (delete…
Anti Slop Fr
>
Optimize Prompt
One deliberate rewrite pass that applies Anthropic's current prompt-engineering best practices to an existing draft prompt (SKILL.md, scheduled-task prompt, API/system prompt, one-off task prompt). Use when the user types /optimize-prompt or says "optimize this prompt" / "improve this prompt" / "tune this prompt" / "apply prompt best practices to X". Input is a pasted prompt, @-mentioned file, or…
Instrument Data To Allotrope
Convert laboratory instrument output files (PDF, CSV, Excel, TXT) to Allotrope Simple Model (ASM) JSON format or flattened 2D CSV. Use this skill when scientists need to standardize instrument data for LIMS systems, data lakes, or downstream analysis. Supports auto-detection of instrument types. Outputs include full ASM JSON, flattened CSV for easy import, and exportable Python code for data engi…
Start
Set up your bio-research environment and explore available tools. Use when first getting oriented with the plugin, checking which literature, drug-discovery, or visualization MCP servers are connected, or surveying available analysis skills before starting a new project.
Scientific Problem Selection
This skill should be used when scientists need help with research problem selection, project ideation, troubleshooting stuck projects, or strategic scientific decisions. Use this skill when users ask to pitch a new research idea, work through a project problem, evaluate project risks, plan research strategy, navigate decision trees, or get help choosing what scientific problem to work on. Typical…
Scvi Tools
Deep learning for single-cell analysis using scvi-tools. This skill should be used when users need (1) data integration and batch correction with scVI/scANVI, (2) ATAC-seq analysis with PeakVI, (3) CITE-seq multi-modal analysis with totalVI, (4) multiome RNA+ATAC analysis with MultiVI, (5) spatial transcriptomics deconvolution with DestVI, (6) label transfer and reference mapping with scANVI/scAr…
Single Cell Rna Qc
Performs quality control on single-cell RNA-seq data (.h5ad or .h5 files) using scverse best practices with MAD-based filtering and comprehensive visualizations. Use when users request QC analysis, filtering low-quality cells, assessing data quality, or following scverse/scanpy best practices for single-cell analysis.
Clinical Trial Protocol Skill
Generate clinical trial protocols for medical devices or drugs. This skill should be used when users say "Create a clinical trial protocol", "Generate protocol for [device/drug]", "Help me design a clinical study", "Research similar trials for [intervention]", or when developing FDA submission documentation for investigational products.
Weckr Integration
Helps developers integrate Weckr, an AI cost and margin tracking SDK, into their application. Use this skill when a user is building an AI feature that calls OpenAI, Anthropic, Gemini, or Kimi and wants to track cost, margin, or spending caps per user, or when a user asks about tracking AI costs, agent loop detection, or per user profitability in their SaaS product.
Weckr Model Pricing
Provides current per-token API prices for large language models across OpenAI, Anthropic, Gemini, and Kimi (Moonshot). Use this skill when a user asks what a model costs, how much input or output tokens cost, which model is cheapest for a task, or wants to compare prices between models or providers. Prices are per million tokens and were verified in mid 2026, so always tell the user to confirm ag…
Weckr Cost Estimator
Estimates what an AI feature will cost per call, per user, and per month before it ships. Use this skill when a user is building or planning an LLM feature and asks how much it will cost to run, whether it is affordable at their price point, how a model choice changes cost, or what per user spend to expect. It reads the feature's code or description, estimates token usage, and projects cost using…
Weckr Margin Audit
Audits whether a SaaS product's pricing plans stay profitable once LLM costs are counted. Use this skill when a user wants to know which of their plans lose money, whether flat or unlimited pricing is sustainable, how much AI cost eats into a plan's margin, or whether their price points cover model spend. It compares each plan's price against the AI cost of a typical and a heavy user and flags pl…
Upgrade
Upgrade an existing project's managed llm-wiki structure without rerunning broad wiki bootstrap. Use when a user asks to migrate, update, or upgrade project-local .llm-wiki scripts, hooks, or changelog structure after installing a newer llm-wiki release.
Wiki Bootstrap
Bootstrap an LLM-maintained wiki for the current project from Pi. Use when a Pi user asks to create, initialize, refresh, or configure a project wiki, LLM wiki, QMD wiki, or codebase knowledge base.
Status
Check whether llm-wiki is current, detect newer marketplace or Pi package versions, and suggest or run the right update command for Claude Code, Codex, or Pi. Use when the user asks about llm-wiki version, plugin/package updates, marketplace refresh, upgrade instructions, project context setup, or whether a new llm-wiki release is available.
Research
Search QMD-indexed project and main cross-project LLM wikis for past patterns, decisions, solutions, and pitfalls before planning or implementation. Use before feature planning, bug fixing, refactoring, architecture work, or code changes when a project wiki may exist.
Wiki Status
Check llm-wiki install status, version freshness, update commands, and project wiki automation from Pi. Use when a Pi user asks about llm-wiki version, package updates, refresh ownership, context setup, or whether a new llm-wiki release is available.
Wiki Research
Search project and main cross-project LLM wikis from Pi for past patterns, decisions, solutions, and pitfalls before planning or implementation. Use before feature planning, bug fixing, refactoring, architecture work, or code changes when wiki context may exist.
Wiki Upgrade
Upgrade an existing project's managed llm-wiki structure from Pi without rerunning broad bootstrap.
Bootstrap
Bootstrap an LLM-maintained wiki for the current project. Use when a user asks to create, initialize, or refresh a project wiki, LLM wiki, QMD wiki, or codebase knowledge base for Claude Code, Codex, or Pi.
Layerkit Fix From Dry Run
Fix failed integration verification from logs, tests, and docs evidence; edit real package files, no invention.
Layerkit Discover Data Layer
Read customer code for domain events/fields; deny secrets; emit domain_spec proposal with source:code.
Layerkit Source Edit Client
Integrate vendors by editing the production datalayer directly; style-match; run the client package verification command.
Layerkit Orchestrate Integration
Master skill — same pipeline for first integrate and contract heal (agent status/next/mark-done).
Layerkit Multi Agent
Coordinate specialist agents for integrate or contract heal; brownfield production code.
Layerkit Privacy Review
Review and update client-owned privacy, consent, hashing, redaction, and egress checks before vendor integration changes ship.
Layerkit Author Map
Draft vendor_map proposal with sources from research evidence; never invent without evidence.
Layerkit Design Flow
Decide the simplest client-owned integration shape from code and vendor evidence; prefer updating existing mappers/adapters.
Layerkit Deletion First
Deletion-first change discipline for Layerkit work. Use before adding code, files, exports, docs, tests, or abstractions; during heal/integration updates; and whenever Codex should prefer modifying or deleting existing code over additive implementation.
Layerkit Research Vendor
Evidence-first vendor research and contract updates (docs/OpenAPI → structured contract → drift vs map); residual human only.
Layerkit Author Processor
Author or update client-side transform/helper code from vendor docs with mandatory citations.
Layerkit Checker Assist
Read-only checker assistant — risk checklist only; never approve, apply, or write checks[].
Layerkit Align Client Style
Analyze existing client package patterns from source evidence; write style notes before source edits.
Layerkit Bootstrap
Install Layerkit into the current repo; seed project store + memory; run doctor.
Layerkit Inventory Surfaces
Inventory package languages/surfaces for multi-lang heals; session file used by Layerkit to enforce all languages updated before PR.
Layerkit Session Handoff
Write a runbook to memory so the next agent can resume the integration loop safely.
Layerkit Design Integration
Choose linear map vs flow (route/foreach/if), multi-step, batch; write design decision to memory.
Analyzing Malware
Analyze suspected malware safely — containment, static triage, sandboxed detonation, unpacking, capability and C2 extraction, IOC production, and YARA rule authoring. Use when handed a suspicious file, hash, or sample, when triaging an alert artifact, or when producing detection content from a specimen.
Triaging Security Alerts
Work a security alert queue to a defensible disposition — separating true positives from false positives and benign true positives, reasoning about base rates before escalating, ordering enrichment by cheapest discriminator, time-boxing, and documenting negative results so a closed alert is evidence rather than a guess. Use when triaging SOC or EDR alerts, deciding whether an alert warrants incid…
Investigating M365 Entra
Investigate a Microsoft 365 and Entra ID compromise where the only evidence is cloud logs — the Unified Audit Log, Entra sign-in and audit logs, OAuth app-consent grants, and mailbox inbox/forwarding rules — to reconstruct business email compromise, token and session theft, and consent abuse. Use for a suspected M365/Entra account takeover with no host or memory to image. Preserve the Unified Aud…
Attacking Active Directory
Attack and enumerate Active Directory environments using Kerberos attacks (Kerberoasting, ASREPRoasting), credential dumping (DCSync, Mimikatz), lateral movement (PtH, PtT), and BloodHound analysis. Use when pentesting Windows domains or exploiting AD misconfigurations.
Analyzing Phishing Emails
Analyze a reported or suspected phishing email safely — parse the Received chain and Return-Path, validate SPF/DKIM/DMARC alignment, extract and defang URLs and attachments, detonate payloads in a sandbox, and pivot on sender infrastructure to produce IOCs and a disposition. Use when a user-reported email, a mailbox artifact, or a `.eml`/`.msg` needs a verdict. Every link is live until proven oth…
Establishing Persistence
Maintain authorized access across reboots and credential changes during red-team post-exploitation — Windows autostart (Run keys, services, scheduled tasks, WMI subscriptions), Linux (cron, systemd, shell profiles, SSH keys), Active Directory (accounts, DCSync rights, tickets), and cloud identity — chosen least-aggressive first, logged for cleanup, and paired with the detection each leaves. Use p…
Writing Yara Rules
Author durable YARA detection rules — meta/strings/condition anatomy, string types and modifiers, structural conditions with file magic and offsets, the PE/ELF/math/hash modules, specificity-vs-durability tuning, atom-aware performance, memory and process scanning, and YARA-X. Use when writing a YARA rule, creating a signature for a malware family or a file/memory artifact, turning IOCs or a capt…
Cracking Passwords
Crack password hashes using hashcat/john, perform password spraying, brute force authentication, and execute pass-the-hash attacks. Use when cracking credentials or performing password-based attacks.
Escalating Windows Privileges
Escalate privileges on Windows systems using service misconfigurations, DLL hijacking, token manipulation, UAC bypasses, registry exploits, and credential dumping. Use when performing Windows post-exploitation or privilege escalation.
Orchestrating Vulnerability Research
Run a sustained, multi-agent vulnerability-discovery campaign against a target — split its attack surface into slices, hunt each slice with a builder agent, and have a separate critic with fresh context adversarially refute every candidate against the real artifact (a reproduced crash, a working request, a proven bypass) before it counts as a finding. Use when tasked to find previously-unknown bu…
Managing Vulnerabilities
Prioritize and drive remediation of a vulnerability backlog by real risk, not raw CVSS — combining severity with exploitation signals (EPSS, CISA KEV), asset exposure and business context, using SSVC-style decisions, distinguishing reachable from merely present, and tracking remediation and exceptions. Use when triaging scanner output, deciding what to patch first, building a risk-based vulnerabi…
Performing Social Engineering
Run authorized social-engineering assessments — pretext development from OSINT, phishing and spearphishing campaigns (Gophish), vishing and smishing, and physical pretexting — to measure the human attack surface and produce awareness-driving metrics. Use when the engagement scope explicitly authorizes testing people, with organizational sign-off and HR/legal coordination. You are testing a consen…
Mapping Attack Techniques
Navigate security work by MITRE ATT&CK tactic and technique — resolve a technique ID or name to the right skill, map a threat intel report or adversary emulation plan to procedures, and run the purple-team loop from technique to detection to validation. Use when a request names an ATT&CK ID like T1003.001, a tactic like lateral movement, an APT group or intel report, or when planning coverage aga…
Engineering Detections
Build, test, and tune detection content — Sigma, YARA, Suricata, and EDR/SIEM queries — mapped to MITRE ATT&CK with explicit false-positive analysis and detection-as-code practices. Use when writing or reviewing a detection rule, converting IOCs or TTPs into alerts, measuring detection coverage, or reducing alert fatigue.
Testing Ics Ot Protocols
Safely assess industrial control system and OT networks and their protocols (Modbus, DNP3, S7comm, EtherNet/IP-CIP, OPC UA, IEC 60870-5-104, BACnet) using a passive-first, safety-gated methodology aligned to the Purdue model and IEC 62443. Use when mapping an OT network, evaluating IT/OT segmentation and zone/conduit rules, or inventorying industrial assets — where a probe that is routine on IT c…
Reviewing Cryptography
Review cryptographic implementations and protocol usage for misuse — weak primitives, nonce and IV handling, key management, authentication of ciphertext, randomness, timing side channels, TLS and JWT configuration, and password storage. Use when auditing code that encrypts, signs, hashes, or authenticates, or when assessing TLS and token configurations.
Producing Threat Intelligence
Turn raw observations, extracted IOCs, and open sources into finished threat intelligence — running the intelligence cycle (direction, collection, processing, analysis, dissemination, feedback), enriching and grading indicators, pivoting on TTPs over atomic IOCs, structuring attribution with calibrated confidence, and packaging tactical/operational/strategic products with TLP and sourcing. Use wh…
Writing Sigma Rules
Author and maintain Sigma detection rules — structure, logsource taxonomy, detection logic with modifiers, false-positive filtering, backend conversion with pySigma, and offline validation with Hayabusa or Chainsaw. Use when translating threat intel into vendor-agnostic detection logic, building a detection-as-code pipeline around Sigma, reviewing or tuning existing Sigma rules, or converting rul…
Testing Web Applications
Test web applications for security vulnerabilities including SQLi, XSS, command injection, JWT attacks, SSRF, file uploads, XXE, and API flaws. Use when pentesting web apps, analyzing authentication, or exploiting OWASP Top 10 vulnerabilities.
Attacking Wireless Networks
Attack WiFi networks using WPA/WPA2 cracking, WPS exploitation, Evil Twin attacks, deauthentication, and wireless reconnaissance. Use when pentesting wireless networks or performing WiFi security assessments.
Threat Modeling
Systematic threat modeling skill for applications, APIs, and systems using STRIDE, PASTA, Attack Trees, DREAD, LINDDUN, and OCTAVE. Use when assessing security architecture, creating data flow diagrams (Mermaid), enumerating threats from OpenAPI specs or architecture docs, building attack trees, mapping threats to NIST/CIS/OWASP ASVS controls, or producing a threat model report. Triggers on reque…
Auditing Mcp Servers
Audit a Model Context Protocol server's own implementation for how it can subvert or exfiltrate from the agent that connects to it — tool-description injection (tool poisoning), tool shadowing and rug pulls, per-tool authorization and input schemas, SSRF via URL-fetching tools, transport and authentication exposure, secrets handling, and toxic tool-flow combinations. Use when the MCP server imple…