mcp-fractera

The DevOps problem is solved. Finally.

You know how to write code. You do not want to think about how to wire Nginx, issue SSL certificates, connect auth to a database, configure PM2, set up reverse proxies, and remember which port does what. Nobody wants to spend a week on that before writing a single line of business logic.

Fractera makes all of that completely automatic. You give it a bare Ubuntu server. It configures everything — Nginx, HTTPS, authentication, database, media storage, five AI coding platforms, an autonomous orchestrator, and private graph memory. You get a fully operational private workspace at http://your-ip:3002 in about 10 minutes. Then you code.

This is not a Docker image you still have to manage. This is not a tutorial you follow step by step. This is a system that deploys itself and gets out of your way.

"Finally someone made it so I don't have to think about deployment, ports, DevOps commands, and all that nonsense."

Quick start — two ways

1. Website (zero terminal)

→ fractera.ai

Enter your VPS credentials. Watch it deploy. Done in ~10 minutes.

2. MCP connector (from your AI chat)

Add to Claude, Cursor, or any MCP-compatible client:

{
  "mcpServers": {
    "fractera": {
      "url": "https://www.fractera.ai/api/mcp"
    }
  }
}

Tell your agent: "Deploy Fractera on my server at IP x.x.x.x" — provide credentials when asked. The agent runs the full deployment and reports back when your workspace is live.

MCP connector: https://www.fractera.ai/api/mcp · Open, no auth
Listed on: Smithery · mcp.so · Glama · Official MCP Registry (io.github.Fractera/deploy)

Production coding + local development — same workflow

Fractera is not just for remote browser coding. It works equally well as a production target for your existing local setup:

1. Connect your Fractera project to a GitHub repository.

2. Develop locally in your IDE with hot reload — your usual flow, zero changes.

3. Push to GitHub when ready.

4. Pull in Fractera workspace and hit Deploy.

Your changes go live in production in minutes. Your own server becomes a self-hosted Vercel: GitHub is the bridge between your local environment and production. Local environment uses the database and file storage that live on your server — no cloud subscriptions.

What you get

Architecture

┌──────────────────────────────────────────────────────────┐
│                    Fractera Workspace                     │
│                                                          │
│  fractera-app              :3000  Shell (Next.js)         │
│  fractera-auth             :3001  Auth (NextAuth v5)      │
│  fractera-admin            :3002  Admin UI (bridges/app)  │
│  fractera-bridge           :3200–3206  AI platform PTYs  │
│  fractera-data             :3300  Media + SQLite          │
│  fractera-rag              :9621  LightRAG memory         │
│  fractera-hermes           :9119  Hermes dashboard        │
│  fractera-hermes-gateway   —      Telegram/messaging      │
└──────────────────────────────────────────────────────────┘

Tech stack: Next.js 16.2 · React 19 · Tailwind v4 · shadcn/ui · SQLite · NextAuth v5 · Node.js · PM2

Real-world use cases

Private team workspace — editors collaborate on content planning in a secure authenticated environment, nothing exposed publicly.

Lead dispatch Kanban — inbound emails from a website form auto-create Kanban cards, routed to field engineers by proximity to minimize travel costs.

Adaptive AI tutor — child completes coding challenges on a public page; parent sees results in a private dashboard and adjusts lessons via Telegram.

Autonomous content loop — agent monitors Telegram channels for trending topics, enriches them via web search, publishes to a blog, and reports traffic stats back to Telegram.

⭐ Star this repo

If Fractera saves you from writing another Nginx config — a star makes a real difference. It helps other developers find the project.

FAQ

Why does Hermes reduce development costs rather than increase them?

There is a widespread belief that an orchestration layer like Hermes is wasteful — that it is better to talk directly to a coding model and skip the middleman. We disagree completely.

The real cost driver in AI-assisted development is not the number of requests — it is context window inflation. Every message in a direct coding session appends to the running context. After two or three hours on a typical project, the context is carrying the full transcript of the session. Most providers charge for both input and output tokens, and because every new message pays for everything before it, costs grow faster than linearly. Real-world data: a Claude Code session on a mid-size project in hour three costs five to ten times what it cost in the first fifteen minutes. When you forget to manage the context window yourself — and everyone forgets eventually — that cost compounds.

Hermes is the external observer that prevents this from happening. It runs on inexpensive models (gpt-5-mini, qwen-turbo, and similar) that cost twenty to fifty times less than frontier coding models. These models do not need sophisticated reasoning — they read a spec, extract what matters, route a task, and immediately clear the context before the next call.

What Hermes does specifically:

• Eliminates session amnesia — without persistent memory, every new session spends the first 10–30% of its tokens re-explaining the project from scratch. LightRAG stores architecture, patterns, anti-patterns, and decisions so the coding prompt is laser-focused from word one.

• Generates precise task specs — instead of a vague "build a dashboard", Hermes reads LightRAG and generates: "In app/dashboard/page.tsx, add a component using the Card pattern from components/ui/card.tsx, following the auth pattern from lib/auth.ts." This eliminates several back-and-forth exchanges per task.

• Clears context immediately — after each task completes, Hermes clears the coding model's context. The next task starts clean, not carrying the entire previous conversation.

• Routes intelligently between models — simple tasks (documentation, refactoring, renaming) go to cheap or free models. Complex architectural decisions go to frontier models. You pay frontier prices only for frontier work.

• Tracks patterns and anti-patterns — Hermes knows what has worked in your project and which approaches caused problems. The coding model does not re-discover the same dead ends.

This is only the surface. The deeper value is in what Hermes prevents: redundant searches for components that already exist, re-solving problems that were solved last week, sending the same context to five models in parallel when one focused call would do. Every one of those is a token cost that never appears on your invoice because it never happened.

We believe the future of efficient AI-assisted development is not about squeezing more out of a single model — it is about a system that knows your project well enough to ask the right model the right question at the right moment, and stop before it becomes expensive. That system is Hermes.

The same AI platforms — yet Fractera ships faster with fewer tokens. Why?

Regular vibe coding puts all the heavy lifting on the AI: design the architecture, write boilerplate, locate the right component, recall what was decided last session. Every token spent on that overhead is a token not spent on your actual feature.

Fractera eliminates that overhead at every layer:

• Production-ready starters — Auth, database, file storage, and routing ship fully pre-configured. The AI skips months of scaffolding and goes straight to your feature from day one.

• Component highlighting — Click any element on your live site to jump directly to its source. No tokens wasted asking "where is the navbar?".

• LightRAG — A persistent vector store that remembers your entire codebase and every architectural decision. Every AI message arrives with full context.

• AI-optimized skills & instructions — Pre-built prompts designed for the right approach on the first try.

• Cross-platform orchestration — All five coding platforms share context. Switching from Claude Code to Gemini CLI does not mean losing the thread.

Tasks that take 10–20 back-and-forth messages in a vanilla AI chat typically resolve in 2–3 focused exchanges inside Fractera.

How does Fractera keep your project stable?

The "free services" formula looks great on day one: Clerk for auth, Supabase for database, a dozen cloud APIs. Then each service quietly updates its pricing. Miss one billing cycle and your project goes dark — no warning, no grace period. Partners who switched to Fractera often share this story: a live project lost because one service expired. By the time they figured it out, the reputation damage was done.

Fractera keeps everything on your server:

• One server, one bill — cost does not scale with your users.

• Pause your business — your data does not disappear. Back it up and restore when ready.

• Code lives on GitHub — recovery is always possible, even if dependencies have aged.

• Built-in AI can help rebuild the project even when some packages are outdated.

Do I pay per token for the AI, or use my own subscription?

You use your own subscriptions — that is the whole point. The five coding platforms (Claude Code, Codex, Gemini CLI, Qwen Code, Kimi Code) run on the subscriptions you already have: you sign in once with your normal account, like a local CLI. No API keys to manage, no per-token billing surprises. This is where the heavy AI work happens, and it costs you nothing beyond the subscription you already pay the AI vendor.

The only per-token piece is small and auxiliary: Brain (Hermes) and Memory (LightRAG) use one inexpensive OpenAI API key — the cheap gpt-5-mini runs at roughly a cent per hour and is plenty to get going, or a Codex subscription if your usage is heavy.

• Coding platforms (Claude Code, Codex, Gemini, Qwen, Kimi) — your existing subscriptions, no per-token billing.

• Brain + Memory — one small OpenAI key (gpt-5-mini, ~1¢/hour), or a Codex subscription if heavy.

• So a Claude Code user pays through their subscription, not per token.

What server specs do I need?

For full AI-coding workloads: 6 cores / 8 GB RAM recommended minimum. Storage depends on your project — 75 GB is a solid baseline.

Once active AI development wraps up, downgrade to 2 cores / 4 GB RAM — typically €1–2 per month.

Can I combine local development with Fractera?

Yes — for developers with an existing local setup, this is often the most natural workflow:

1. After setting up your Fractera project, connect it to a GitHub repository and push the initial codebase.

2. Pull the code to your local machine. Develop the way you always have — your IDE, hot reload.

3. When ready to ship, push your changes to GitHub.

4. Return to your Fractera workspace, pull from GitHub, and hit Deploy.

Your changes go live in production in minutes. Your own server becomes a self-hosted Vercel: GitHub is the bridge between your local environment and production. Your local environment uses the database and file storage that live on your server — no cloud subscriptions required.

Can I bring my existing project into Fractera?

Yes. Connect your existing GitHub repository to your Fractera workspace and start coding with AI immediately. Depending on your project's complexity, some initial migration steps may be needed — Fractera's built-in AI assistants can guide you through.

Can I deploy a finished project to Vercel instead?

Yes — once a project is complete and no longer needs active AI-assisted development, you can export it to Vercel. Keep in mind: the AI coding workspace, terminal access, LightRAG memory, and all five development platforms only run on your own server. Vercel and cloud storage pricing can also escalate quickly under real-world traffic. Migrating back is straightforward — your code is already on GitHub.

Can I deploy a project without AI?

Yes — you can deploy a plain server with no AI at all. A common choice among experienced developers who want to offload cloud resource management and avoid DevOps overhead. The server can be synced with your local IDE (VS Code, etc.) or simply used to host existing projects on infrastructure you control.

• Optional authentication — turn it on whenever you need it.

• Local database and object storage — included, no cloud subscriptions.

• Freedom from cloud lock-in — meet regulatory requirements and avoid surprise bills.

During deployment, switch to custom mode and select which services you want.

Do I need a Claude Code subscription to deploy via the MCP connector?

Not necessarily. If the Fractera connector is your only custom MCP, you can use it on the free tier — no subscription required. If you already use another custom MCP, either remove it before adding the Fractera connector, or upgrade to a paid Claude plan (Claude Pro, ~€20/month) to keep multiple connectors at once. Note: this is Anthropic's own plan — it is unrelated to Fractera, which is always free.

Step by step — what does the full path from buying a server to shipping a product look like?

1. Buy a VPS (Contabo recommended). Receive your IP, login, and password.

2. Enter credentials on the website or via the MCP connector. Watch real-time deployment events. Two emails arrive: start and completion.

3. Open the admin panel link from your completion email. Sign up with a real email and password — the first user becomes the Architect (admin).

4. Have your own domain? Open Settings → Personal Domain, copy the A-record, set it at your DNS provider.

5. Authorize coding platforms (Claude Code, Codex, Gemini CLI, Qwen Code, Kimi Code). One-time browser-based OAuth — same as local CLI.

6. Talk to Brain in the built-in web chat — it opens automatically and is the main way you use the system; just add a model to start. Optionally add a Telegram bot token to also chat from your phone.

7. Tell Hermes what to build: "Let Claude do X and Codex do Y" — both run in parallel through orchestration.

8. Upload documents, requirements, and business context to Company Brain (LightRAG) — or ask Hermes to automate ingestion.

9. Your site is live from minute one. After every development cycle, it updates automatically — no manual deploy step.

Regional FAQ

Does Fractera comply with Russian data residency requirements (152-ФЗ)?

Yes — the public side of your project remains fully under Russian jurisdiction. Database, file storage, and authentication run on your server; deploy it at a Russian provider (e.g. Timeweb Cloud) and your users' personal data never leaves Russia. Note: Fractera provides the infrastructure that enables compliance — the final responsibility depends on your chosen hosting and how you design data processing in your application.

AI agents are Western. Does that conflict with import substitution?

No. AI-assisted development is the private side: you write code with any assistant and can work from anywhere. The AI is not part of the production data processing pipeline for your users, so it does not conflict with 152-ФЗ. If full sovereignty matters — use Qwen Code and Kimi Code (Chinese models, not subject to Western restrictions).

How does Fractera help save tokens?

The main source of savings is LightRAG project memory: it stores the context of your codebase and decisions, so AI agents do not need to re-read the project from scratch every time. The more memory knows, the fewer tokens go to "remember what's where." On top of that, Hermes orchestration and local MCPs reduce 10–20 message tasks to 2–3 exchanges.

Roadmap & vision

When you start a business for the first time, almost everything is unfamiliar: registering the company, standing up a website, finding traffic, balancing unit economics, and scaling. An AI agent helps at every step. Along the way it proposes solutions — some right, some wrong; some save money, some waste it. Until now this zero-to-product experience lived only in your head. Now it can be digitized.

The AI that solved every task alongside you, listened to your conversations, prepared your presentations — can assemble not a step-by-step manual but a living knowledge base. Not yes/no, but the complex conditional logic of decisions that modern AI can analyze and carry over to a different region, with different inputs and a different outcome.

This is what we call the product loop — and it is the product we are building next.

Do you have a partner program?

Yes. You recommend Fractera deployment, your readers buy a VPS from any host with an affiliate program, the host pays you directly. Fractera provides the infrastructure (mirror page, embed widget, MCP connector, partner cabinet), the host pays the commission.

Direct affiliate programs (RuVDS, Timeweb, Hostinger, Hetzner) approve in a day and pay up to 30–40% of the customer's first payment. Network programs (Contabo via CJ.com) pay a fixed amount per new customer.

→ Partners page · Questions: admin@fractera.ai

Regional partnerships — we are looking for co-founders, not resellers

Fractera is infrastructure — and infrastructure is only as good as the support around it. We are actively looking for official regional partners to build localized versions of the platform together: adapted for local regulations, delivered in the native language, and backed by people who understand the regional market from the inside.

This is not a reseller arrangement. A regional Fractera deployment can mean your own branded product, your own customer base, and your own revenue stream — built on an open-source foundation that is already production-ready. The technical work is done. What a regional partner brings is market knowledge, local trust, and the ability to serve customers in a way that a remote team cannot.

If you are in a market where cloud data residency matters, where documentation needs to be in a language other than English, or where developers simply need someone local to talk to — that is exactly the gap we want to fill together.

→ Regional Partners · admin@fractera.ai

Links

• Website & deploy: fractera.ai

• Knowledge base for AI agents: fractera.ai/mcp-info

• MCP connector: https://www.fractera.ai/api/mcp

• Contact: admin@fractera.ai