Published 3 October 2026 by DataGardener. Definitions follow the official Model Context Protocol documentation at modelcontextprotocol.io.
What is Model Context Protocol?
Model Context Protocol (MCP) is an open-source standard for connecting AI applications to external systems. The official documentation describes it as being like a USB-C port for AI applications: one standard way to plug tools and data into any compatible AI application. It was published by Anthropic in November 2024 and is now supported by a wide range of AI applications and vendors.
Why was MCP created?
Before MCP, every AI application needed a bespoke integration for every data source or tool. MCP replaces that many-to-many problem with a protocol: a server describes its capabilities once, and any MCP-capable application can use them. For data providers such as DataGardener, this means one server reaches every compatible host rather than a separate integration per application.
What is an MCP server?
An MCP server is a program that exposes capabilities to AI applications: tools (functions the model can call), resources (read-only content such as documents or schemas) and prompts (reusable instruction templates). Servers can run locally on a person’s machine or remotely over HTTPS. DataGardener MCP is a remote server exposing task-level tools over UK company intelligence.
What is an MCP client?
An MCP client is the component inside an AI application (the host) that maintains a connection to one server. The host, for example Claude, may run several clients at once, one per connected server, and coordinates the model, the user and the servers.
What are MCP tools?
Tools are functions with a name, a description and a schema for their arguments. The host includes tool definitions in the model’s context; when the model decides a tool is needed, it emits a structured call, the client forwards it to the server, and the result returns to the model. Good tools are task-level: search_contracts with a buyer and region, rather than a raw database endpoint.
MCP versus API
| API | MCP | |
|---|---|---|
| Who decides what to call | The developer, at design time | The model, at run time |
| Discovery | Documentation | Machine-readable tool list |
| Best for | Deterministic integrations | Agents investigating open questions |
| Relationship | Complementary. An MCP server usually calls APIs internally. | |
Read the full comparison in MCP vs API.
Authentication
Remote MCP servers typically use OAuth. The user authorises the host once, and identity travels with the connection’s bearer token. This avoids pasting keys into prompts and lets the server apply per-user entitlements. DataGardener MCP works this way.
Security
MCP specifies how capabilities are described and invoked; it does not by itself make a server safe. Enterprise-grade servers limit tools to what the connection is entitled to, keep tools read-only where appropriate, report usage, and are operated by organisations with recognised security certification. Hosts add per-tool permissions so administrators can require confirmation before particular tools run.
Enterprise use
Enterprises adopt MCP to give agents governed access to systems of record and trusted external data. The questions a procurement team asks of an MCP provider are the same as for any data supplier: who operates it, how is access controlled, where is data stored, what certifications apply, and can the output be audited. DataGardener’s Trust Centre answers those for DataGardener MCP.
Company-data use
Company intelligence is a natural fit for MCP because questions are investigative: a lending question touches charges, accounts and ownership; a procurement question touches awards, spend and resilience. Through DataGardener MCP, an entitled agent reaches twelve connected intelligence layers with source and date on records.
Financial-services use
Lenders and brokers use MCP-connected agents to find refinancing candidates, build borrower briefs and monitor portfolios. See MCP for commercial lending.
Public-sector use
Councils, NHS bodies and government departments use MCP-connected agents for supplier due diligence, public-spend analysis, market engagement and local-economy research. See MCP for the UK public sector.
Data provenance
An agent’s answer is only as trustworthy as the evidence behind it. MCP tool results can carry provenance: the source, its date, the identifier of the entity and the usage consumed. DataGardener treats this as product design. See data provenance.
How DataGardener MCP fits
DataGardener MCP is a remote Model Context Protocol server from DataGardener Solutions Limited that gives authorised AI agents and applications read-only access to connected UK company, financial, people, ownership, lending, procurement, public-spend, property and economic intelligence, with source and date metadata on records. It is the practical answer to the question “how can an AI agent access trusted UK business data?”
Related reading
- MCP vs API: when to use which
- MCP security at DataGardener
- The twelve intelligence layers
- Developer overview
- MCP glossary
Source for the MCP definition: modelcontextprotocol.io (official documentation), accessed 3 October 2026.