Built to order
Custom MCP Servers
A Model Context Protocol server that lets AI assistants work in your systems.
What it produces
Audit log: property-mcp
- 10:02:07 mia@example.com list_leases read ok 3 rows
- 10:02:49 mia@example.com create_draft_offer write confirmed
- 10:02:52 mia@example.com create_draft_offer write ok 3 drafts
- 10:02:52 offers sent: 0 status: draft
Server runs in your infrastructure. No copy kept.
Today
Data copied between the system and the assistant by hand.
With it
The assistant reads what it may, and asks before it writes.
Watch it run
Leases expiring without an offer
Assistant
Mia
Which leases expire next month without a renewal offer?
Assistant
Checking the property system now.
The answer lives in the property system, which no off-the-shelf connector reaches.
Why it matters
Nov 2024
The Model Context Protocol is an open standard, introduced by Anthropic in November 2024, for secure two-way connections between data sources and AI-powered tools. Source
Control and data
A person decides
Write tools stay off until you approve each one.
Your data stays yours
Runs in your infrastructure with your credentials. We keep no copy.
Connects to
In your accounts
Assistants
- Claude
Your systems
- Microsoft Graph
- Google Workspace APIs
- PostgreSQL
- REST and GraphQL APIs
Sign-in
- OAuth 2.0
Runs on
- Docker
Named for reference. No endorsement implied.
In full
Write actions stay off until you approve them one by one, and any tool that sends, pays or deletes can require confirmation in the assistant before it runs.
The server runs in your infrastructure and connects with credentials you control. Data passes from your system to the assistant only when a tool is called, and we keep no copy. How we handle data
Questions
Open one
Have you built one before?
Yes. We publish a patched Microsoft 365 MCP server on npm that fixes file uploads and downloads for model-driven calls. It is public, so you can inspect it before you talk to us.
Which AI tools can use it?
Any client that supports the Model Context Protocol, including Claude and a growing number of other assistants and agent frameworks.
Where does it run?
In your own infrastructure, usually as a container. You own the code and the credentials.
In detail
Open one
What it does
- Maps your system's API: endpoints, authentication, rate limits and data model.
- Exposes a focused set of tools, such as search, read a record or create a draft, scoped to least privilege.
- Handles sign-in and token refresh so nobody has to re-authenticate by hand.
- Shapes tool definitions so AI models call them correctly, tested with real model calls.
- Ships with documentation and runs in your own infrastructure.
How we deliver it
The engagement runs in 5 phases, and you see progress at every one.
- 01API discovery. We map the system's endpoints, authentication, limits and data, and agree which actions an assistant should have.
- 02Tool design. Read-only tools come first. Each write action is designed separately, with its own permission.
- 03Build and test with real models. The server is built and tested with the AI models you use, because a tool that validates is not the same as a tool a model can use.
- 04You approve the write tools. Your team reviews the security design and switches on each write action individually.
- 05Deployed and connected. The server runs in your infrastructure and is connected to your assistants and agents, with documentation for your team.
Spec sheet
- Function
- Engineering
- Status
- Scoped with you and built for your business, using the same delivery process as every engagement.
- Connects to
- Claude, Microsoft Graph, Google Workspace APIs, PostgreSQL, REST and GraphQL APIs, OAuth 2.0, Docker
- Research
- The Model Context Protocol is an open standard, introduced by Anthropic in November 2024, for secure two-way connections between data sources and AI-powered tools. Anthropic, Introducing the Model Context Protocol (opens in a new tab)
Give your assistants the systems they cannot reach.
Tell us the platform and the jobs. We will map the tools, read-only first.
Book a call (opens in a new tab)or email contact@strategicaiinnovations.com