← Services

MCP Integration

Your existing app, usable by AI.

We wrap the systems you already run in an MCP server — scoped tools, real interfaces via MCP Apps, and an integration that works in Claude, ChatGPT, Cursor, and your own platform.

What you get

  • An MCP server over your existing API, database or app
  • Interactive MCP Apps widgets, not just text replies
  • Runs in Claude, ChatGPT, Cursor and self-hosted platforms
  • Scoped tools with a server-side authorization boundary
  • You own the server, the widgets and the specs

Best for Teams with a working product that AI assistants cannot reach yet

From Free discovery, then €85/h — scoped by integration complexity

Start a conversation

Request a callback

Leave your details and a named senior engineer gets back to you within one business day. No sales script, no obligation.

By sending this, you agree to our privacy policy. We use your details only to reply.

What this is

Your systems already work. The gap is that an assistant cannot reach them: it can discuss your orders, invoices, or stock levels in the abstract, but it cannot look one up, and it must not change one blindly. MCP integration closes that gap without touching what runs in production today.

The Model Context Protocol is the standard interface between AI hosts and the systems they act on. We put an MCP server in front of your existing REST API, GraphQL endpoint, database, or desktop application, expose a deliberate set of tools, and — where it earns its place — give those tools a real interface with MCP Apps. Nothing is rewritten. The integration lives beside your application, not inside a rewrite of it.

Three ways we integrate an existing application

Which shape fits depends on how your application is deployed and who needs to reach it. Discovery ends with that decision written down.

Most common

Wrapper server

A separate MCP server in front of what you already expose.

  • Speaks to your existing REST, GraphQL or SQL surface
  • Zero changes to the running application
  • Fastest path to a working integration
  • Tool surface designed, not auto-generated from endpoints

Embedded mode

MCP as a second mode of your own binary.

  • The application itself answers as an MCP server
  • One codebase, one deployment, no translation layer
  • The binzaar pattern: one Go binary, three modes
  • Best when we also own or maintain the application

Remote server

Hosted over streamable HTTP for many users.

  • OAuth-authenticated, per-user permissions
  • One deployment serves a whole team or product
  • Runs on your infrastructure or ours, EU regions
  • Monitoring, versioning and rollout handled properly
MCP Apps: interfaces in the chat, not walls of text

A tool that returns JSON forces the assistant to describe your data in prose. A tool that returns an interface lets your people act on it. When the assistant lists open invoices, it can hand back a sortable table with per-row actions; when it needs structured input, a validated form the user submits straight back as a tool call.

This is a standard, not a trick that breaks on the next host update. In January 2026, Anthropic and OpenAI standardised interactive UI as MCP Apps (SEP-1865) — the first official extension to the Model Context Protocol, carried by the io.modelcontextprotocol/ui namespace. A widget built once renders in any compliant host: Claude, ChatGPT, VS Code, Cursor, Postman, or the private workspace we build for you.

Our open-source library gadget is how we ship those interfaces. It generates table, form, card, and card-list widgets as self-contained HTML resources embedded in the server binary — themed to the host, formatted to the host’s locale and timezone, locked down by a strict Content Security Policy, with no external CDN dependency at all.

A validated form rendered by gadget inside an assistant conversation Change Ada's role to admin Here is the record — edit and submit: Message the assistant
A form widget inside the thread: typed fields, client-side validation, server-side error mapping. Submitting it calls the MCP tool directly — no copy-paste back into the chat.
A responsive grid of record cards rendered by gadget inside an assistant conversation Show me the users I can edit Four users. Filter, sort, or act on any card: Message the assistant
The same data as a card grid, with filtering, sorting and per-item actions. Built once, rendered by whichever host your team uses.
Proof: the MCP servers we ship in the open

We point at working software. Every pattern on this page is one we run in our own repositories, under the MIT licence, in public.

binzaar

Open source

A single-binary local app store: browse a GitHub-hosted catalog, install release binaries with SHA-256 verification, update, and scaffold new microapps. One binary runs as a terminal UI, an MCP stdio server, and an HTTP catalog server — the embedded-mode pattern described above, including MCP Apps widgets.

GobbolttviewMCP stdioMCP Apps
github.com/Techthos/binzaar

gadget

Open source · pre-release

The Go library behind our MCP Apps interfaces: table, form, card and card-list widgets emitted as self-contained HTML, SDK-agnostic at the core with an adapter for the official MCP go-sdk. Compliant with the io.modelcontextprotocol/ui extension, revision 2026-01-26.

GoTypeScriptesbuildMCP Apps (SEP-1865)
github.com/Techthos/gadget

clockwork

Open source

Our flagship dual-mode time tracker: an MCP server that lets an assistant log work in natural language and turn git commits into worklogs, plus a keyboard-driven terminal UI over the same data.

GoMCP stdio
github.com/Techthos/clockwork
The standards we build to

Building on the published specification is what keeps your integration portable. A bespoke assistant plugin is tied to one vendor’s host; an MCP server is not.

MCP core

Tools, resources and prompts modelled properly — with stdio for local processes and streamable HTTP for remote deployments.

MCP Apps (SEP-1865)

The official UI extension, io.modelcontextprotocol/ui, revision 2026-01-26 — the same spec our gadget library implements.

Host compatibility

Claude, ChatGPT, Cursor, VS Code and Postman, plus self-hosted Open WebUI and LibreChat workspaces.

Language and runtime

Go where a single distributable binary wins, Python or TypeScript where your team has to maintain it afterwards.

Security and what the assistant is actually allowed to do

The reason most internal systems are still not connected to an assistant is not capability, it is permissions. We treat the tool surface as an authorization boundary, not a prompt.

Read-side first

Write operations ship after the read path is proven, each behind an explicit scope and, where it matters, a confirmation step in the widget.

Real authentication

Per-user OAuth on remote servers, service credentials on local ones. Never one shared admin token standing in for every caller.

Enforced server-side

Every tool checks who is calling and what they may touch in the server, where a prompt cannot argue with it. Instructions in a document are not permissions.

Sandboxed widgets

MCP Apps widgets are self-contained documents under a strict CSP with no external requests — nothing phones home from inside your conversation.

Auditable calls

Tool invocations are logged with caller, arguments and result status, so you can answer what the assistant did and when.

EU by default

EU data residency, signed DPAs, disclosed sub-processors — the same posture as every other Techthos engagement.

How we engage & pricing

Per-milestone and stop-anywhere, billed by effort at €85/hour. What an integration costs depends on its complexity — how many tools, how clean the existing interface is, whether write operations and interactive widgets are in scope — so we scope it after Discovery rather than guessing at a fixed price.

  1. Discovery Free
    We read your API or schema, agree which hosts must reach it, and deliver a written tool map with a complexity verdict
  2. Integration spec By effort · €85/h
    The tool surface on paper: each tool, its arguments, its permission, its widget — usable to brief any team · about 1 week
  3. Pilot server By effort · €85/h
    A working MCP server with the first tools, running in your host, read-side only · 1–2 weeks
  4. Full integration By effort · €85/h
    The complete tool set, write operations, MCP Apps widgets, authentication, deployment, CI and handover · 2–8 weeks
  5. Maintenance from €70 / mo
    Optional retainer: spec revisions, host compatibility, and the small changes that follow

Billed by effort at €85/hour; the total follows integration complexity and is fixed per SOW after the free Discovery. You own the server source, the widget definitions, the specifications and the deployment. Stop after any milestone — the tool map alone is a usable artifact.