Skip to content
Custom Connectors
SolutionsConnectorsPricingResourcesBlog
Build my connector

How does a custom Claude connector get built?

A done-for-you process in three plain steps: scope, build, use. We do the technical work and the ongoing maintenance, so your team just uses Claude inside the software you already run.

Build my connectorBrowse connectors

Key takeaways

  • The process is three steps: scope your tool, build a hosted MCP connector, then use Claude in your day-to-day.
  • Read and write tools are kept separate; create and update flow freely while delete and send wait for your confirmation.
  • OAuth authenticates as your own account, so the connector never has more access than you already have.
  • On-prem and firewalled systems are supported through secure connectivity such as IP allowlisting.
  • We maintain the connector as APIs, schemas, and tokens change, so it keeps working over time.

How does the build process work?

The build process works in three steps: scope, build, use. First we scope your tool and the actions Claude should take on a short call. Then we build and host a custom MCP connector with OAuth and scoped read-write tools. Finally you use Claude in your normal conversations while we maintain it.

  1. 01

    Scope

    On a short call we map your tool, what Claude should be able to read and do, which write actions need a confirmation step, and how authentication works for your system. You leave the call with a clear picture of the exact connector we will build.

  2. 02

    Build

    We build and host a custom MCP connector with read tools and write tools kept separate, OAuth wired to your existing login, and scoped permissions. For on-prem or firewalled systems we sort out secure connectivity, including IP allowlisting, before anything ships.

  3. 03

    Use

    You add the connector in Claude and use it in your normal conversations. Claude reads, drafts, creates, and updates inside your tool, while delete and send wait for your confirmation. We maintain the connector as your software and its API change over time.

How long does a connector take to build?

How long a connector takes to build depends entirely on scope. A single-tool connector against a well documented API moves faster than a multi-tool build or an on-prem system that needs secure connectivity set up, since each added action and integration point takes more work. We confirm the window with you on the scoping call, once the scope is clear.

What do you need from me?

We need three things from you: one short scoping call, a way to authenticate (usually OAuth against an account that already has the access you want Claude to use), and a sense of which write actions matter most. For firewalled systems we also agree on secure connectivity. Everything else, including the build and hosting, sits with us.

How do you handle on-prem or firewalled systems?

On-prem or firewalled systems are handled with secure connectivity agreed during scoping. The most common approach is IP allowlisting: your network only accepts traffic from our known hosting address, and the connector reaches your internal system through that approved path. The connector keeps the same scoped, separated read-write design as any cloud build.

How do you keep write actions safe?

Write actions stay safe through scoped tool separation and human-in-the-loop gating. Read and write tools are distinct, so create and update flow freely inside a conversation while delete and send require your confirmation first. The connector inherits your account permissions, so Claude never gets more access to a system than you already have.

Frequently asked questions

Do I need to be technical to use a custom connector?
No. We design, build, host, and maintain the connector for you, so there is no MCP server to run and no code for your team to manage. Once it is live you add it in Claude and use plain language. The technical work, from OAuth to schema updates, stays on our side.
How does authentication work, and is it secure?
The connector uses OAuth to authenticate as your own account, so it inherits the exact permissions you already have. It never grants Claude more access than your account holds. It also acts only inside a Claude conversation you start, and destructive or outbound actions are gated behind a confirmation you control.
Can the connector both read and write, or just read?
Both. We build read tools and write tools as separate, clearly scoped capabilities. Claude can read records, create them, and update them freely inside a conversation. Higher-risk actions like delete and send are gated behind a human-in-the-loop confirmation, so nothing irreversible happens without your explicit approval.
Why not just use the built-in or native Claude connector?
Built-in connectors are a great starting point, but each ships a fixed, generic set of tools scoped the same way for every user. They often stop short of the specific actions, custom fields, or workflows your team needs. The built-in Gmail connector, for instance, drafts a reply as a new message rather than inside the original thread. We build a custom connector that delivers the full capability and gates destructive or outbound actions behind your approval.
What if my system sits behind a firewall or VPN?
We handle on-prem and firewalled systems regularly. During scoping we agree on secure connectivity, which usually means IP allowlisting our hosting so your network only accepts traffic from a known source. The connector then reaches your internal system through that approved path, with the same scoped read-write design as any other build.
What happens when my software updates and the API changes?
That is what the maintenance plan covers. APIs change, schemas shift, and OAuth tokens expire, and we keep the connector working through all of it. When your tool adds fields or endpoints you want Claude to use, we extend the connector. You get a connector that keeps working, not one that breaks after the next update.
How involved is my team during the build?
Lightly. We need one scoping call, access or an account to authenticate against, and a few answers about which write actions matter most. After that the build is on us. We come back to confirm scope and walk you through adding the connector in Claude, but day-to-day building does not sit on your team.

Tell us what you run. We'll build the connector.

Pick your tool, choose what Claude should do, and get an instant scope and price. No call required.

Build my connectorTalk to us