[{"data":1,"prerenderedAt":618},["ShallowReactive",2],{"home":3,"article-\u002Fblog\u002Fwhat-is-mcp-model-context-protocol-explained":318},{"id":4,"title":5,"articleCta":6,"capabilities":11,"capabilitiesSectionText":52,"capabilitiesSectionTitle":53,"closing":54,"contact":59,"engagements":71,"engagementsSectionText":103,"engagementsSectionTitle":104,"extension":105,"faq":106,"footer":144,"founder":149,"getStartedSectionTitle":164,"getStartedSteps":165,"heroCtaButton":178,"heroKicker":179,"heroPoints":180,"introSectionText":184,"introSectionTitle":185,"meta":186,"microcopy":187,"navbar":203,"processSectionText":226,"processSectionTitle":227,"processSteps":228,"qualifier":245,"services":264,"servicesSectionText":285,"servicesSectionTitle":286,"stem":287,"trustBar":288,"whyUs":295,"whyUsSectionTitle":316,"__hash__":317},"home\u002Fhome.yaml","AI Automation, Agents & Software Architecture Consultancy",{"kicker":7,"title":8,"primary":9,"secondary":10},"Working on something like this?","A 30-minute call tells us both whether we can help.","Book a discovery call","More field notes",[12,16,20,24,28,32,36,40,44,48],{"title":13,"icon":14,"description":15},"Custom Software, Built Around Your Workflow","code","One-size-fits-all fits nobody in particular. We design and ship software that matches the way your team operates, not the way a generic product wants you to.",{"title":17,"icon":18,"description":19},"AI Automation & Agentic Workflows","workflow","n8n orchestrations, LangGraph agents, and RAG pipelines that quietly handle the repetitive work in the background. Fewer tickets, fewer copy-paste jobs, more throughput.",{"title":21,"icon":22,"description":23},"MCP Integrations for AI-Ready Tools","plug","We build Model Context Protocol servers that expose your internal systems to Claude, ChatGPT, and Copilot, so your team can use AI safely against the data that matters.",{"title":25,"icon":26,"description":27},"ERP, CRM & Data Integrations","database","We wire up the systems that should have been talking years ago. Clean data flows between ERP, CRM, and operational tools, plus dashboards that turn it into decisions.",{"title":29,"icon":30,"description":31},"Full-Stack Engineering","layers","Frontend, backend, databases, APIs: we cover the whole build in Vue, React, Node, Python, PostgreSQL, Redis and MongoDB. One team owning the stack top-to-bottom means fewer seams for bugs to hide in.",{"title":33,"icon":34,"description":35},"Cloud, Platform Engineering & DevOps","cloud","AWS, Azure, GCP. Kubernetes, Terraform, CI\u002FCD pipelines, infrastructure as code. We build the platform your app runs on so deploys are boring, environments are reproducible, and on-call isn't a horror story.",{"title":37,"icon":38,"description":39},"Security, SecOps & Compliance","shield","Security built in from day one, not bolted on before the audit. Threat modelling, least-privilege access, secret management, dependency hygiene, and the paper trail to prove it. Comfortable working under HIPAA, SOC 2, and enterprise security review.",{"title":41,"icon":42,"description":43},"Architecture & System Design","network","Monolith, microservices, event-driven or serverless: we pick the shape that matches your scale, your team, and your budget, not the one that looks impressive in a talk. Clear boundaries, clean contracts, documentation a human can read.",{"title":45,"icon":46,"description":47},"Industry Experience That Transfers","building","Healthcare (HIPAA, HL7, FHIR), insurance, fintech, logistics, SaaS. Different industries share the same underlying problems, and the patterns that solve them carry over.",{"title":49,"icon":50,"description":51},"The Idea You Haven't Built Yet","lightbulb","Got a weird integration, a backlog of 'we should automate this', or a product idea that no vendor covers? Tell us about it. The interesting work tends to start with a messy whiteboard.","When the SaaS subscriptions stop adding up and the spreadsheets start creaking, it's time for something built around how your business works. Here's where we come in.","When Off-the-Shelf Software Stops Fitting",{"kicker":55,"title":56,"accent":57,"tagline":58},"Next step","Bring the problem.","We'll bring the architecture.","Software architecture and AI automation consultancy, led by a senior architect.",{"title":60,"description":61,"form":62},"Let's Talk About the Work You'd Rather Not Do Manually.","Tell us what's slow, brittle, or eating your team's week. We'll tell you whether automation, an AI agent, or a proper architecture rethink is the right call, and what it would take.",{"title":63,"description":64,"namePlaceholder":65,"emailPlaceholder":66,"messagePlaceholder":67,"submitButton":68,"errorMessage":69,"thankYouMessage":70},"Get in Touch","Tell us about your project or pain points.","Your Name","Your Email","How can we help?","Let's Talk","Oops! Something went wrong. Please try again later, or email us at","Thank you! We'll be in touch shortly.",[72,78,83,88,93,98],{"title":73,"featured":74,"bestFor":75,"description":76,"price":77},"Discovery Sprint",true,"Best for: teams with a fuzzy problem and a real deadline.","A focused one-to-two week engagement to turn a vague brief into a concrete plan. You walk away with an architecture recommendation, a costed roadmap, and a clear go\u002Fno-go.","Fixed fee · scoped up front",{"title":79,"bestFor":80,"description":81,"price":82},"Architecture Review","Best for: existing systems, vendor proposals, or pre-build sanity checks.","We review your codebase, infrastructure, or an incoming quote and give you a plain-English report: what's solid, what's risky, what to fix first, and what it'll cost.","Fixed fee · 1–2 weeks",{"title":84,"bestFor":85,"description":86,"price":87},"System & Architecture Audit","Best for: due diligence, pre-acquisition, or deep technical forensics across a product suite.","A multi-week, independent deep-dive across architecture, codebase, CI\u002FCD, security posture, and cloud-readiness. Delivered as a structured written report for executive stakeholders: technical risks, modernization gaps, and prioritized recommendations, backed by evidence.","Fixed fee · 3–12 weeks",{"title":89,"bestFor":90,"description":91,"price":92},"Fractional Architect Retainer","Best for: founders and SMEs without a full-time technical lead.","Ongoing senior technical leadership on a monthly retainer. You get a CTO-level adviser on call for hiring, architecture, vendor calls, and the decisions that are expensive to get wrong.","Monthly retainer · flexible scope",{"title":94,"bestFor":95,"description":96,"price":97},"Fixed-Scope Build","Best for: automations, integrations, and well-defined features.","You know what you need. We scope it, quote it, and ship it in four to twelve weeks, with working software demoed every fortnight.","Fixed quote after discovery",{"title":99,"bestFor":100,"description":101,"price":102},"Ongoing Delivery Partner","Best for: longer builds and teams that want a steady senior partner.","A small, senior pod working alongside your team on a rolling basis. Great for modernization, platform work, and anything that needs architectural consistency over months, not weeks.","Monthly · capped effort","Pick the engagement shape that fits the problem. We'll tell you which one you probably need.","Ways to Work With Us","yaml",{"title":107,"subtitle":108,"cta":109,"questions":110},"Frequently Asked Questions","The questions we hear most often on discovery calls.","Still have questions? Email us at",[111,114,117,120,123,126,129,132,135,138,141],{"question":112,"answer":113},"What does Innovi Pro do?","We're a small software consultancy. Two things we do well: help you make the big technical decisions (architecture, stack, vendors, scaling), and build the software and automations that come out of those decisions. Think senior engineering partner rather than body-shop agency.",{"question":115,"answer":116},"Who do you work with?","Founders and SMEs who've outgrown spreadsheets and SaaS duct tape, scale-ups that need a senior hand on the architecture, and established companies that want to modernize operations without a year-long transformation project. Industry-wise we've shipped work in healthcare, insurance, fintech, logistics, and SaaS.",{"question":118,"answer":119},"How is this different from hiring an agency or a dev shop?","You talk to the people doing the work. No layered account managers, no pitch-deck team handing you off to juniors. Engagements are led by a senior software architect, delivered by a small trusted network, and sized to the outcome — not to fill a bench.",{"question":121,"answer":122},"Do you only do AI work?","No. AI and automation are a big part of what we do because that's where most of the quick wins are right now, but the core of the business is solid software architecture and custom development. Sometimes the best answer is a better database schema, not an AI agent.",{"question":124,"answer":125},"How do engagements typically start?","With a short discovery call: no deck, no hard sell. We want to understand what is slow, brittle, or blocked. From there you get one of three things: a fixed-scope project proposal, a retainer for ongoing architectural guidance, or an honest 'you don't need us for this' and a pointer in the right direction.",{"question":127,"answer":128},"How much does a project cost?","It depends on scope and on how involved you want us to be. Discovery sprints, reviews and audits are fixed-fee and scoped up front. Builds get a fixed quote once discovery has defined them. Advisory and delivery partnerships run on a monthly retainer. Every proposal lists the deliverables, what is out of scope and how each deliverable will be accepted. We'd rather quote accurately than quote fast.",{"question":130,"answer":131},"How quickly can you start?","Discovery calls usually happen within the week. For smaller automation or advisory work we can often start within two weeks. Larger builds are planned around your timeline and ours. We deliberately keep the active client list small so the people you meet are the ones actually doing the work.",{"question":133,"answer":134},"Will you work with our existing team and tools?","Almost always, yes. Most of our work augments an in-house team rather than replacing it. We adapt to your stack, your processes, and your codebase; we're not here to rewrite everything in our favorite framework.",{"question":136,"answer":137},"What about security, compliance, and data protection?","Taken seriously by default. We've worked under HIPAA in healthcare and inside the security expectations of large regulated enterprises. NDAs are standard, data stays where it should, and we're comfortable operating in environments where audit trails and access control aren't optional.",{"question":139,"answer":140},"Do you have a preferred automation platform?","We use n8n heavily because it provides the right balance of power and transparency for most business automation. That said, we're not religious about tools. If a project is better served by a custom workflow engine, a LangGraph agent, or even plain code, that's what we'll recommend.",{"question":142,"answer":143},"We're not sure what we need yet. Is it too early to talk?","That's the best time to talk. A 30-minute call before you commit to a direction tends to save months of rebuilding later. Bring the problem, not a spec, and we'll help you figure out the shape of the solution.",{"title":145,"address":146,"email":147,"copyright":148},"Innovi Pro","Cluj-Napoca, Romania","contact@innovi.pro","© 2025 Innovi Pro. All rights reserved.",{"kicker":150,"name":151,"role":152,"photoAlt":153,"bio":154,"tags":157},"The architect behind Innovi Pro","Mihai Farcas","Lead Software Architect","Mihai Farcas, Lead Software Architect at Innovi Pro",[155,156],"I help teams across industries design and deliver custom software that is secure, scalable, and built around real operational needs. My work combines **architecture consulting** with hands-on delivery through a trusted network of engineers, so you can ship without hiring a full team first.","I write about the engineering behind this work on [mihai.ltd](https:\u002F\u002Fmihai.ltd\u002Fblog) and on the [Innovi Pro blog](\u002Fblog).",[158,159,160,161,162,163],"Software architecture","Agentic AI","n8n","MCP","Kubernetes","Healthcare IT","Get Started In 4 Easy Steps",[166,169,172,175],{"title":167,"description":168},"Reach Out","Email us or fill out the contact form describing your project requirements.",{"title":170,"description":171},"Discovery Call","We'll schedule a discovery call to discuss your vision and goals.",{"title":173,"description":174},"Get a Quote","Get a tailored proposal outlining the recommended solutions, timeline, and cost.",{"title":176,"description":177},"We Start Building","Once you're ready, we kick off the development process and keep you updated every step of the way.","See How We Can Help","Software architecture · AI automation · Cluj-Napoca",[181,182,183],"Led by a senior software architect","Production n8n & agentic AI systems","HIPAA, HL7 and FHIR experience","We build the AI agents, n8n workflows, and MCP integrations that take the repetitive, unglamorous work off your team, so people spend their time on the parts of the job that need a human.","AI Automation Built Around [Your Business.]",{},{"primaryCta":9,"stackTitle":188,"scorecardLink":189,"partnerBadge":190,"qualifierCta":193,"ideaCta":194,"featuredEngagement":195,"engagementsHelp":196,"engagementsHelpLink":197,"whyUsCard":198,"formNote":202},"The stack we ship on","The method",{"kicker":191,"label":192},"Production stack","n8n automation","Sounds like us? Let's talk","Tell us about it","Typical first step","Not sure which one fits?","Book a call and we'll tell you.",{"kicker":199,"title":200,"cta":201},"Start here","A 30-minute call. No deck, no hard sell.","Book it","NDAs are standard. Your message goes straight to the architect.",{"items":204},[205,208,211,214,217,220,223],{"title":206,"url":207},"Services","\u002F#services",{"title":209,"url":210},"Capabilities","\u002F#capabilities",{"title":212,"url":213},"Process","\u002F#process",{"title":215,"url":216},"Engagements","\u002F#engagements",{"title":218,"url":219},"Why Us","\u002F#why-us",{"title":221,"url":222},"Blog","\u002Fblog",{"title":224,"url":225},"Book a Call","\u002F#contact","No mystery, no 40-page Statements of Work. Here is what working with us looks like.","How We Work",[229,233,237,241],{"title":230,"meta":231,"description":232},"1. Discovery Call","30 min · free","A 30-minute conversation. You describe what's slow, broken, or blocking. We ask the awkward questions. No deck, no pitch, no obligation. If we're not the right fit, we'll say so and point you at someone who is, when we know one.",{"title":234,"meta":235,"description":236},"2. Architecture & Scoping","1–2 weeks · fixed fee","A short, paid deep-dive of one to two weeks. You get a written architecture recommendation, a risk list, and a costed roadmap. This alone is often the most valuable thing clients buy from us.",{"title":238,"meta":239,"description":240},"3. Build, Automate, Integrate","2-week cycles","Delivered in short, visible iterations: two-week cycles with working software at the end of each. We work in the open: shared backlog, shared repo, direct access to the engineers.",{"title":242,"meta":243,"description":244},"4. Handover & Keep-The-Lights-On","Optional retainer","You get clean code, documentation a human can read, and a team that's trained to own it. We stick around on a light retainer if you want ongoing architectural guidance, or step fully out. Your call.",{"title":246,"subtitle":247,"items":248},"Is this you?","Most of our best engagements start with one of these.",[249,252,255,258,261],{"title":250,"description":251},"Your SaaS stack has quietly become a spreadsheet with extra steps.","Ten tools, three of them half-used, and the real process still lives in someone's inbox. You need software that fits the way your team works.",{"title":253,"description":254},"You inherited a codebase and you're not sure if it's salvageable.","Maybe a previous agency, maybe a co-founder who left. You need a clear-eyed second opinion before you put another cent into it.",{"title":256,"description":257},"Your team spends half its week on work a computer should be doing.","Invoicing, triage, copy-pasting between systems, answering the same five questions. You're ready to automate but want it done properly, not with a fragile chain of scripts.",{"title":259,"description":260},"You're about to commit to a 6-figure build and want a second pair of eyes.","An independent architecture review before you sign. Vendor quotes decoded into plain English. A roadmap you can follow.",{"title":262,"description":263},"You know AI should be helping, but you don't know where to start.","Not another chatbot demo: automations and agents plugged into the systems you already run, with straight advice about what is worth doing and what is hype.",[265,275],{"title":266,"description":267,"icon":268,"deliverables":269},"Software Architecture & Technical Consultancy","A senior architect in your corner, without the full-time headcount. Independent system and architecture audits, technical due diligence, vendor code reviews, cloud and container readiness assessments, and costed modernization roadmaps. We produce written, data-driven reports executives can act on and engineers actually respect. Especially useful before acquisitions, major rebuilds, or any six-figure technical decision.","architect",[270,271,272,273,274],"Architecture and system audits","Technical due diligence","Vendor code reviews","Cloud and container readiness","Costed modernization roadmaps",{"title":276,"description":277,"icon":278,"deliverables":279},"AI Automation, Agents & n8n Workflows","Production-grade automation, not demo-ware. We design n8n workflows, build custom agentic AI with LangGraph, and ship MCP servers that plug your internal tools straight into Claude, ChatGPT, and Copilot. Autonomous agents for support, invoicing, and back-office ops. Clean integrations between ERP, CRM, and AI that keep working on a Monday morning.","automation",[280,281,282,283,284],"n8n workflows","LangGraph agents","MCP servers for Claude, ChatGPT and Copilot","Support, invoicing and back-office agents","ERP and CRM integrations","Two ways to work with us: strategic consultancy when the decisions are expensive, and hands-on delivery when it's time to build. Often both.","What We Do","home",{"title":289,"items":290},"Senior experience you can reach directly.",[291,292,293,294],"10+ years shipping enterprise software","Experience across healthcare, insurance, fintech & SaaS","Published on DEV, YouTube & industry blogs","HIPAA · HL7 · FHIR experience",[296,300,304,308,312],{"title":297,"icon":298,"description":299},"A Team Led by an Architect","compass","Every engagement is led end-to-end by a senior software architect: the same person on the discovery call, in the design doc, and on the pull requests. No handoff to a junior delivery team, no layer of account managers translating between you and the build.",{"title":301,"icon":302,"description":303},"Deep n8n Expertise, Serious About AI","badge","Extensive production experience in n8n, LangGraph, MCP, and the wider agentic AI stack. We pick the right tool for the job instead of forcing every problem into the same framework.",{"title":305,"icon":306,"description":307},"Enterprise Experience, Startup Pace","gauge","Patterns learned on large-scale enterprise systems, applied at the speed a smaller team needs. Secure, scalable, and not over-engineered.",{"title":309,"icon":310,"description":311},"Small Team, Direct Access","users","You talk to the people doing the work. Decisions happen in hours, not change-request cycles. Lean on purpose, so your budget goes into outcomes instead of overhead.",{"title":313,"icon":314,"description":315},"AI-Accelerated Delivery","sparkles","We use AI to move faster where it helps (code scaffolding, research, testing) and keep humans in charge of the parts that matter. You get more software, sooner, without the usual quality trade-off.","Why Teams Work With Us","6q6dgWrLG8hN7zDaK4pfIxGN1megf6_hGCU5JE2fAnQ",{"id":319,"title":320,"author":145,"body":321,"canonical":600,"date":601,"description":602,"extension":603,"image":349,"imageAlt":604,"imageFallback":605,"meta":606,"navigation":74,"ogImage":605,"path":607,"published":74,"readingTime":608,"seo":609,"stem":610,"tags":611,"twitterCard":615,"type":616,"__hash__":617},"blog\u002Fblog\u002Fwhat-is-mcp-model-context-protocol-explained.md","MCP Explained: The Standard That Makes AI Agents Actually Useful",{"type":322,"value":323,"toc":584},"minimark",[324,328,332,340,343,351,356,359,378,381,388,391,395,398,404,410,416,436,439,443,446,451,454,457,461,464,467,471,478,481,485,488,494,500,506,509,513,516,530,533,544,547,551,554,557,571,575,578,581],[325,326,320],"h1",{"id":327},"mcp-explained-the-standard-that-makes-ai-agents-actually-useful",[329,330,331],"p",{},"An LLM on its own is an expensive autocomplete. It can write you a poem about your database, but it cannot query it, change it, or tell you what broke at 3 a.m. The thing that turns a language model into a useful agent is the set of tools you give it. For most of the last two years, wiring up those tools was a custom job for every single app.",[329,333,334,335,339],{},"The ",[336,337,338],"strong",{},"Model Context Protocol"," (MCP) is the fix. Anthropic open-sourced it in late 2024 and it has become the default way to plug tools, data, and prompts into any LLM-powered application. If you have heard MCP described as \"USB-C for AI,\" that is actually the right metaphor: one connector, many devices, no drivers to write.",[329,341,342],{},"This article explains what MCP is, the three pieces that matter, why it changes how you build AI products, and where it hurts.",[329,344,345],{},[346,347],"img",{"alt":348,"src":349,"title":350},"The Model Context Protocol architecture: host application and MCP client on the left, MCP server in the middle, external tools, resources, and prompts on the right.","\u002Fblog\u002Fdiagrams\u002Fmcp-architecture.webp","MCP in one picture: host, client, server, tools.",[352,353,355],"h2",{"id":354},"the-problem-mcp-solves","The problem MCP solves",[329,357,358],{},"Before MCP, integrating a tool into an AI app looked like this:",[360,361,362,366,369,372,375],"ul",{},[363,364,365],"li",{},"Write a custom function definition in OpenAI's format.",[363,367,368],{},"Write a different custom definition in Anthropic's format.",[363,370,371],{},"Re-write it again for LangChain.",[363,373,374],{},"Re-write it again for your internal agent framework.",[363,376,377],{},"Keep four copies in sync forever.",[329,379,380],{},"Every model vendor shipped its own tool-calling schema. Every framework wrapped it differently. And every single integration (Jira, Postgres, GitHub, your CRM) got re-implemented in every codebase that touched it. Multiply by every team in your company and you have a combinatorial mess.",[329,382,383,384,387],{},"MCP collapses this. You write the integration ",[336,385,386],{},"once",", as an MCP server. Any MCP client can use it without you doing anything extra: Claude Desktop, Cursor, your own app, even a competitor's app. The model does not care whether the server is written in Python, TypeScript, or Rust. The client does not care what tools the server exposes until it asks.",[329,389,390],{},"That is the whole pitch. It is the same reason USB-C won. Not because it is the most elegant connector ever designed, but because it is the same connector everywhere.",[352,392,394],{"id":393},"the-three-pieces","The three pieces",[329,396,397],{},"MCP has exactly three moving parts. Learn these and the rest is details.",[329,399,400,403],{},[336,401,402],{},"1. The host"," is the application the user actually interacts with. Claude Desktop, Cursor, Zed, a custom chatbot, an internal ops console. The host embeds the LLM and owns the UI.",[329,405,406,409],{},[336,407,408],{},"2. The client"," lives inside the host. It speaks MCP on behalf of the host. Each host can run many clients, one per connected MCP server, and the client handles the JSON-RPC back-and-forth so the host does not have to.",[329,411,412,415],{},[336,413,414],{},"3. The server"," is where the interesting stuff lives. An MCP server exposes three kinds of things:",[360,417,418,424,430],{},[363,419,420,423],{},[336,421,422],{},"Tools:"," functions the model can call. \"Create a Jira ticket.\" \"Run this SQL.\" \"Send this email.\"",[363,425,426,429],{},[336,427,428],{},"Resources:"," read-only data the model can pull in. A file, a database schema, a Confluence page.",[363,431,432,435],{},[336,433,434],{},"Prompts:"," pre-written prompt templates the server offers to the host. \"Summarize this incident in the house style.\"",[329,437,438],{},"The client-server handshake is boring and that is the point. The client asks the server what it can do. The server answers. The model picks from that menu. The server does the work and returns the result. Same shape every time, same wire format every time.",[352,440,442],{"id":441},"what-changes-when-you-adopt-it","What changes when you adopt it",[329,444,445],{},"Three things change, and they all compound.",[447,448,450],"h3",{"id":449},"you-stop-writing-glue-code","You stop writing glue code",[329,452,453],{},"In a pre-MCP world, a sizable fraction of any AI codebase is adapters, one function per tool per framework. With MCP, every integration is a server you write once. The servers are small; most are a few hundred lines. And because the protocol is public, there is already a growing ecosystem of community servers for GitHub, Slack, Postgres, Google Drive, Stripe and more.",[329,455,456],{},"For our own tooling we now default to \"write the MCP server first, the UI second.\" The server becomes the canonical contract. Any client, including whatever LLM app we build next, gets it without extra work.",[447,458,460],{"id":459},"your-agent-stops-being-a-silo","Your agent stops being a silo",[329,462,463],{},"Classic agent frameworks bake tool definitions into the agent's code. The agent only knows what the dev hard-coded. MCP flips this: the agent discovers tools at runtime from whichever servers are connected. Add a new server, the agent can immediately use it. No redeploy, no code change on the agent side.",[329,465,466],{},"This matters more than it sounds. It means the product team can ship new capabilities by standing up servers, not by shipping new agent releases. It means users can bring their own MCP servers. It means you can build an agent that does not know what it will be asked to do tomorrow and still copes when the request arrives.",[447,468,470],{"id":469},"security-finally-has-a-natural-seam","Security finally has a natural seam",[329,472,473,474,477],{},"Because every tool call goes through a server, the server is the perfect place to enforce auth, rate limits, audit logs, and scoping. Want the model to query Postgres but only the ",[14,475,476],{},"analytics"," schema, only SELECT statements, only for this tenant? That is a single policy in one server, not forty policies spread across every prompt in your stack.",[329,479,480],{},"For regulated work (healthcare, finance, anything with PII), this is the first time the security seam has been in a sane place. You can put your zero-trust gateway in front of the MCP server and every LLM, every client, every host inherits the controls.",[352,482,484],{"id":483},"where-mcp-still-hurts","Where MCP still hurts",[329,486,487],{},"It is not a finished standard. Three things to know before you bet a platform on it.",[329,489,490,493],{},[336,491,492],{},"Auth is under-specified."," The spec added OAuth in early 2025 but the flows are still awkward for server-to-server scenarios. Most production deployments roll their own token handling. Expect churn here over the next year.",[329,495,496,499],{},[336,497,498],{},"Long-running work is ugly."," MCP is request-response. If a tool call takes 30 minutes, say training a model or running a scraper, you have to model it as \"start job\" + \"poll status.\" Doable, but the protocol does not help you.",[329,501,502,505],{},[336,503,504],{},"Observability is DIY."," The protocol says nothing about tracing. If you care about which tool got called with what arguments and how long it took (you do), you have to instrument every server yourself. OpenTelemetry fits well; we add it by default.",[329,507,508],{},"None of these are blockers. They are the usual early-standard pain points and the community is moving fast on all three.",[352,510,512],{"id":511},"when-to-use-it","When to use it",[329,514,515],{},"Use MCP when any of these are true:",[360,517,518,521,524,527],{},[363,519,520],{},"You are building an agent that needs more than two tools.",[363,522,523],{},"You expect the set of tools to grow, especially if non-engineers will add them.",[363,525,526],{},"You need the same set of integrations to work across multiple AI apps.",[363,528,529],{},"You are in a regulated industry and need a clean enforcement point for tool access.",[329,531,532],{},"Skip it when:",[360,534,535,538,541],{},[363,536,537],{},"You are shipping a single prompt with a single tool and no intent to grow.",[363,539,540],{},"Your use case is pure generation, with no tool calls and no data retrieval.",[363,542,543],{},"You are inside a framework that already does this well and you have zero plans to leave it.",[329,545,546],{},"Most production AI work falls into the first bucket. That is why MCP adoption has been so fast.",[352,548,550],{"id":549},"how-we-use-mcp-at-innovi-pro","How we use MCP at Innovi Pro",[329,552,553],{},"Every agent we build for clients now starts as an MCP server plus a thin host. The server owns the domain: clinical workflows for healthcare, deal data for sales tools, infra state for internal ops. The host is whatever UI the customer needs: Slack, a web app, a VS Code extension.",[329,555,556],{},"Two recent patterns we keep reaching for:",[360,558,559,565],{},[363,560,561,564],{},[336,562,563],{},"One MCP server per bounded context."," If you have four domains, you get four servers. Each one is owned by the team that owns the domain. This mirrors microservices discipline and scales much better than a \"god server\" with everything in it.",[363,566,567,570],{},[336,568,569],{},"Readonly resources before tools."," Start by exposing data as MCP resources. Let the model read first. Add write tools only once you see what it actually needs. You end up with a smaller, safer surface.",[352,572,574],{"id":573},"the-bottom-line","The bottom line",[329,576,577],{},"MCP is not magic. It is a narrow, well-scoped protocol that makes integrations cheap and composable. That is it. But because integrations are the single biggest cost center in real AI products, making them cheap changes what you can build.",[329,579,580],{},"If you are starting a new AI feature today, start with an MCP server. If you have an existing agent, the migration we have seen takes days, not weeks. If you are on the buying side, picking AI tooling for your company, \"speaks MCP\" is a question worth asking.",[329,582,583],{},"The tools your AI can reach are the product. MCP is how you give it more of them, without the wiring eating your budget.",{"title":585,"searchDepth":586,"depth":586,"links":587},"",2,[588,589,590,596,597,598,599],{"id":354,"depth":586,"text":355},{"id":393,"depth":586,"text":394},{"id":441,"depth":586,"text":442,"children":591},[592,594,595],{"id":449,"depth":593,"text":450},3,{"id":459,"depth":593,"text":460},{"id":469,"depth":593,"text":470},{"id":483,"depth":586,"text":484},{"id":511,"depth":586,"text":512},{"id":549,"depth":586,"text":550},{"id":573,"depth":586,"text":574},"https:\u002F\u002Finnovi.pro\u002Fblog\u002Fwhat-is-mcp-model-context-protocol-explained","2026-04-15","The Model Context Protocol is the USB-C of AI agents. Here is what MCP is, how the client-server-tool model works, and why it turns an LLM into a product.","md","Architecture diagram of the Model Context Protocol showing the flow from host application to MCP client to MCP server to external tools, resources, and prompts.","\u002Fblog\u002Fdiagrams\u002Fmcp-architecture.png",{},"\u002Fblog\u002Fwhat-is-mcp-model-context-protocol-explained","8 min",{"title":320,"description":602},"blog\u002Fwhat-is-mcp-model-context-protocol-explained",[161,338,612,613,614],"AI agents","LLM integration","software architecture","summary_large_image","article","NRY1GGnrQB64RTQPEqdLvalOsTuRiyWd-QQ36uWMn18",1791637290285]