What is MCP?
MCP stands for Model Context Protocol. Instead of building a custom integration for every AI + tool pairing, MCP provides a single, standardised protocol that lets an AI agent “plug in” to many tools and services consistently.
Think of it like a USB port for AI, a universal adapter that acts as a centralised gateway between your Agent and the tools it needs to use.
What this means for you
Reduced engineering effort, Integrate an MCP server once, and Toqan can plug into it. No need to build bespoke connectors for every tool.
Faster time to value, Because the interface is standardised, adding new systems becomes significantly faster.
Scoped access & governance (coming soon), MCP supports permission-based access, so you grant access to specific actions rather than handing over wholesale system access.
Connector marketplace (coming soon), Publish or use MCP connectors in a marketplace, lowering the barrier for non-technical users to connect their tools.
Key concepts
MCP runs on a client–server architecture:
- MCP Client, the component that connects to an MCP server and retrieves context. In this case, that’s Toqan.
- MCP Server, describes the actions it supports (e.g. “list pull requests,” “get table data”) and translates those high-level requests into actual API calls or database queries.
Connecting Toqan to your systems
Traditionally, integrating Toqan with tools like Databricks, GitHub, or internal systems required custom engineering, writing code, building APIs, handling maintenance. With MCP, that changes:
- Toqan acts as the MCP client and talks to MCP servers you host
- You host MCP servers for each system you want Toqan to access (Databricks, GitHub, internal tools, etc.)
- You don’t have to wait, if you have the infrastructure and technical capacity, you can deploy MCP servers today and start connecting
Security checklist
MCP servers can handle sensitive data. Before connecting anything, go through this list:
Prioritise official MCPs, always start with MCPs provided by the platform itself (Databricks, Datadog, GitHub, etc.). These are the safest option.
Use trusted registries carefully, if no official MCP exists, community options are available in: , Official MCP Registry, community-owned, backed by Anthropic, GitHub, and Microsoft , Glama.ai MCP Directory, reputable third-party directory, but not official
If you’re unsure whether a provider can be trusted, ask your internal team before deploying.
Always self-host, never connect to an MCP server hosted by a third party online. Deploy and host the MCP on your own infrastructure, following standard security practices (containerisation, access controls, etc.).
Need help setting up? Ask Toqan directly, it can guide you through the process.