Skip to content
Custom Connectors

Claude connector development, done for you

A custom connector puts your software inside Claude: claude.ai, the Desktop app, or a Team plan where every teammate connects as themselves. You scope the actions in the build wizard, we build and host the connector, and Claude reads and acts in your tool with your approval on anything irreversible.

Lives inside Claude

No new app to learn. The connector shows up in the Claude your team already uses, and plain conversation drives it.

Per-user OAuth

On a Team plan, each person connects as themselves, so Claude acts with that person's permissions and nothing more.

Approval-gated writes

Create and update flow in conversation, while delete and send wait for a confirmation the user controls.

Built for your account

Scoped to your fields, pipelines, and workflows, not a generic template of what the tool usually looks like.

Key takeaways

  • A custom Claude connector appears inside the Claude your team already uses: claude.ai, Claude Desktop, or a Team plan.
  • On Team plans each person authenticates through OAuth as themselves, so Claude acts with that person's real permissions.
  • Create and update happen in conversation; delete and send wait for an explicit approval.
  • As of July 2026, 62 percent of 264 business tools have no Claude connector, so for most software custom development is the only route in.
  • We host and maintain the connector after go-live, including migrations as the MCP spec Claude builds on evolves.

What does a custom connector look like inside Claude?

A custom connector shows up as a set of named tools Claude can use in any conversation. You ask about a customer and Claude reads the live record. You ask for an update and Claude writes it into the tool. There is no separate app or dashboard: the interface is the conversation you were already having.

Because the connector is built for your account rather than a generic template, Claude sees your custom fields and pipelines and uses your vocabulary for them. The case studies show what that looks like in practice, from in-thread Gmail drafting to a heavily customized Pipedrive.

How does authentication work on a Claude Team plan?

On a Team plan the connector is shared, but the access is not: each teammate authenticates through OAuth as themselves. Claude then acts with that person's real permissions in the tool, actions are attributed to the right person in the tool's own history, and any individual grant can be revoked without touching the rest of the team.

That per-user model is the difference between a connector a company can actually roll out and a shared service account nobody wants to answer for. The full trust model, including scoping and revocability, is on the security page.

How do write actions stay safe inside Claude?

Write actions stay safe through separation and approval gating. Read and write tools are built as distinct, scoped capabilities, so create and update flow inside a conversation while delete and send stop and wait for your confirmation. Claude proposes the action, shows you what it will do, and runs it only after you approve.

Is Claude connector development different from MCP development?

A Claude connector is an MCP server, so the engineering overlaps, but the service is different. Claude connector development is opinionated about the Claude experience: how tools are named so Claude picks the right one, how Team-plan OAuth is set up, where approval gates sit, and how the connector is added and rolled out in your Claude.

If you are comparing the underlying engineering, pricing, and timelines, the MCP server development page covers the productized build itself, and managed MCP server hosting covers how it stays running afterward. For where custom connectors fit next to Claude's built-in directory, see Claude integrations. And if your tool is one of the many without a listing, that gap is exactly what our coverage study of 264 business tools measures: as of July 2026, 62 percent have no Claude connector at all.

How do I get one built?

Start in the build wizard: pick your tool, choose what Claude should read and do, and see the scope and fixed price immediately, from $5,000 plus $500 per month for hosting and maintenance. Most single-tool builds go live in about two weeks. Prefer to talk it through first? Book a scoping call instead.

The step-by-step process is on the how it works page, and the full tier breakdown is on the pricing page.

Frequently asked questions

Which Claude plans work with a custom connector?
We build for the Claude your team actually uses: claude.ai in the browser, the Claude Desktop app, and Team plans where the whole company shares the connector. During scoping we confirm your plan and set the connector up accordingly, and go-live includes connecting it inside your Claude and walking your team through it.
How does a Team plan rollout work?
The connector is added once, then each teammate authenticates through OAuth as themselves. That per-user model matters: Claude inherits each person's real permissions in the tool, actions land in the tool's history under the right name, and revoking one person's grant does not touch anyone else's access.
What does using the connector actually look like?
Like a normal Claude conversation. You ask about a deal, a ticket, or an order, and Claude reads the live record. You ask it to draft, create, or update, and it does so inside the tool. When a step is irreversible, like delete or send, Claude proposes the action and waits for your approval before running it.
Is this different from the connectors in Claude's directory?
Yes. Directory connectors ship a fixed, generic tool set scoped the same way for every customer. A custom connector is built against your account: your custom fields, your pipelines, the specific write actions your team needs, with gating placed where you want it. And for the many tools with no directory listing at all, custom is the only path.
Can one connector cover several tools?
A connector maps to one tool, but a workflow can span several: we build a connector per tool and the orchestration that runs them as one flow, so a trigger in one system drives the right actions in the others. The build wizard prices multi-tool workflows the same transparent way as single builds.
Who maintains the connector as Claude and my tool evolve?
We do. The monthly plan covers hosting, monitoring, OAuth token health, fixes when your tool's API changes, and migrations as the MCP specification that Claude connectors are built on keeps evolving. The managed MCP server hosting page describes exactly what that covers month to month.

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.