About the project
CoreSynth.io is a hosting platform built around the idea that an AI agent should not stop at support chat. It should help operate real infrastructure: provisioning instances, inspecting logs, assisting with diagnostics and handling selected operational actions in a controlled, auditable way.
The product is aimed at teams that want managed hosting with deeper technical capability behind the UI — not just a dashboard and the rest handled manually over SSH. According to the internal project materials, the platform is currently positioned as an active product in closed beta.
What it solves
A lot of hosting products handle ordering well, but fall apart when real operations begin. Backups, node-level debugging, process inspection and network troubleshooting still end up in the hands of an admin doing manual work.
CoreSynth is designed to move that layer into an agent runtime with explicit tools and controlled write actions. The project materials directly reference capabilities such as VM operations, log access, provisioning and DNS updates.
Architecture
The platform is split into three layers with separate failure domains:
- Control panel and billing — the commercial and user-facing layer built with Next.js and custom PHP billing
- Core nodes — the Proxmox infrastructure running LXC/KVM workloads
- Agent runtime — an isolated MCP-based tool layer for operational actions
That separation matters. The web layer, billing layer and compute layer are not fused into a single runtime, which makes the system more resilient and easier to evolve.
Tech stack
WEB → Next.js 15 + Tailwind 4
API → Node 22 + tRPC + Zod
BILLING → PHP 8.2 + custom billing for CZ market needs
DATA → PostgreSQL 15 + Redis
INFRA → Proxmox VE + LXC + KVM + Ansible
EDGE → NGINX + Fail2ban + CrowdSec
AGENT → Claude / OpenAI + MCP tool registry
AI OPS → local OpenAI-compatible services over OllamaThere is also a verified internal service running on the server called CoreSynth Embedding API. It exposes OpenAI-compatible embedding endpoints, runs as a systemd service and uses a local Ollama model. That makes it a strong technical signal that CoreSynth is supported by real production-style AI infrastructure, not just product messaging.
Key capabilities
- Agent-assisted provisioning — moving from order flow to a running instance without a fully manual handoff
- Operational diagnostics — working with logs, VM state and infrastructure tools through the MCP layer
- Controlled actions — write operations are designed as explicit, supervised actions rather than hidden automation
- Custom billing for the Czech market — invoices, PDFs and local tax logic as part of the platform
- Internal AI services — local embedding endpoints and other support components operated as part of the stack
Who it is for
CoreSynth makes the most sense for teams that want to:
- run their applications without building a full internal ops department,
- have more than a simple hosting panel,
- reduce friction around provisioning and daily operational work,
- combine app infrastructure, billing and tooling into one product.
Outcome
From a portfolio perspective, CoreSynth is strong because it connects two worlds that are often kept separate: modern product engineering and low-level infrastructure operations. The result is a platform where Next.js, PHP billing, PostgreSQL/Redis and Proxmox/Ansible coexist with an agent layer designed for real execution.
It is a product project, but also an infrastructure and systems-design project — which is exactly what makes it interesting.
What’s next
The project materials mention a longer-term direction around an MCP-native agent, wider self-service onboarding and partner-facing APIs. Regardless of roadmap scope, the current shape is already clear: hosting where AI is part of day-to-day operations, not just a UI add-on.