Coming soon

MCP Fabric is coming to market soon. The platform is still being completed. This site is here so you can explore the product and get in touch.

Talk to us

Platform capabilities in detail.

Each page describes one part of MCP Fabric: what it is for, how you use it, and whether it is core or turned on separately.

Interactive chat

A guest writes on the website. If it is a booking, the chat asks for the details, puts an offer together, and after they agree it goes to a person. One click finishes the reservation. Live for guesthouse lodging at Penzion Pod Farou (podfarou.cz).

Agents

A useful agent needs more than an AI model. It needs a clearly defined job, the right information, access to the tools it needs and boundaries it must not cross. MCP Fabric lets you create specialized agents and connect them to business knowledge and processes so each agent has a clear purpose instead of becoming another general-purpose chatbot.

MCP

Your team may use Claude, Codex, Cursor, Windsurf or other MCP-compatible clients. You should not have to build the same business integration for each one. MCP Fabric provides a shared layer between AI clients and business capabilities, so the same tool or process can be used safely from multiple places.

Identity

Connect MCP Fabric to your organization's identity so access reflects real people, roles and membership. Users do not need a separate account simply to use the AI platform, and access rules can align with the way your organization already manages identities.

Monitoring

MCP Fabric includes a production operations stack: metrics, centralized logs and Grafana dashboards. You can see how services, traffic and automated processes behave — and you keep product audit as a separate control, not a screenshot of a chart.

Automated processes

An automated process is more than a diagram on a canvas. It defines what happens, in what order, under which conditions and when a person needs to step in. You can combine information retrieval, decisions, business-system actions, approval and delivery of the result into one repeatable process. Once designed, the AI can run that process consistently.

Capabilities

Interactive chat

A guest writes on the website. If it is a booking, the chat asks for the details, puts an offer together, and after they agree it goes to a person. One click finishes the reservation. Live for guesthouse lodging at Penzion Pod Farou (podfarou.cz).

Monitoring

MCP Fabric includes a production operations stack: metrics, centralized logs and Grafana dashboards. You can see how services, traffic and automated processes behave — and you keep product audit as a separate control, not a screenshot of a chart.

Knowledge search

Connect documents to search, agents, chat and automated processes. The model sees the passages that matter — not the whole knowledge base — so you save tokens. Knowledge stays with the organization that uses it.

Process diagram

An automated process defines what happens, in what order, under which conditions and when a person should step in. What you draw is what runs.

Automated process steps

From starting points and agents to approvals, actions and notifications. The company calendar is one of those steps — an appointment in the calendar, not an engineering shortcut.

SEO monitor

seo-monitor-service runs optional on-page checks and rank tracking (GSC / SERP), Seznam Webmaster indexation, reports and notification e-mails. It is gated by the organization licence.

Probe

probe-service stores profiles and runs (including ZAP and k6 style jobs) behind a licence gate. Artefacts stay on the probe volume, not in the automated process process diagram.

Website checks

A guest or a colleague writes. The team changes the website. Someone says yes. The pages go live. Then a real browser opens them, walks the path a visitor would, and brings a screenshot back to the team. That is website checks: the last look before you trust the change.

Datastore

Searchable text lives in the database. Photos and PDFs are separate — browse them in admin and use drop-off links.

Connectors

The platform composes tools, capabilities and connections: MCP, REST, SQL, GitHub, SSH, Ansible, Terraform, WinRM/PSRP and other deployed connections. Calls go through organization policy and sensitive credential references.

Ansible

Ansible connections bind inventory and playbooks into governed automated process steps. A run uses an approved playbook, sensitive credential-referenced credentials and organization policy. Chat does not get a free ansible-playbook shell.

Ansible workspaces

Versioned playbook trees, published revisions and starter scenarios for a connection.

SSH

SSH connections store host, port, user and keys in the vault. Automated processes run approved command profiles. A free-form remote command from the model is off; a named profile is used instead.

Terraform

A Terraform connection points at modules you have approved. Automated processes can validate, plan, apply and show the result. Destroy stays off unless the connection allows it and a person has approved. Typical work: a Proxmox VM, deploying or upgrading a stack, preparing an environment. Provider credentials — including Proxmox API tokens — stay in the organization vault. The run can be local or over an existing SSH connection.

WinRM / PSRP

WinRM and PowerShell Remoting (PSRP) are one Windows connection family. You store the host and choose WinRM or PSRP, keep passwords and certificates in the vault, then test the path or run approved Windows Ansible books. Same idea as SSH: a governed connection, not a shell in the request.

VPN

VPN connections (IPsec / IKEv2, SSTP or WireGuard, client or server role) sit in the same governed catalog as SSH and Ansible. From a server connection in admin you generate secrets once, then download ready profiles and install commands grouped by iPhone, Android, Windows and Ubuntu. Typical path: gateway first, then SSH, playbooks or Windows remoting on the far side.

GitHub

A GitHub connection uses a personal access token or a GitHub App stored as credentials. Development and operations processes call the GitHub API through that connection, for your organization only.

Data connections

Data connections and datasources bind SQL, REST and GraphQL endpoints. Validate after save, then expose operations as automated process tools. Connection strings never belong in the process diagram.

Forms and webhooks

A customer submits a form — and instead of a silent inbox, your process runs. You keep the address, the expected fields and the history in admin. You can pause it, roll it back or replay a received request. Email, Teams and WhatsApp can start the same kind of process.

Email

Connect a mailbox (classic IMAP or Microsoft 365). A new message starts a process, attachments are stored, and they do not belong in the model’s prompt. The same kind of process can start in Teams or on WhatsApp.

Teams

People already write in Teams. An incoming message can start or continue a automated process — with the same access rules and sensitive credentials as email or a form. First you connect the organization, then only allowed senders get through.

WhatsApp

A customer writes on WhatsApp Business and the message can start the same process as an email or a form. This is official WhatsApp Business, not a generic social chat.

Notifications

notification-service (and related customer notification flows) delivers outbound messages used by SEO reports, operational alerts and automated process steps. SMTP and channel sends use sensitive credential-referenced credentials.

Ready to move beyond the chatbot?

Build AI that understands your business, works with your knowledge, follows your processes and can actually finish the job. You decide which actions are allowed, when a person must step in and how much AI usage your organization can consume.

Request a demo