Skip to content
Betav0.62.0
All articles

Set up the MCP gateway and choose your first connectors

Give every computer's AI tools one governed MCP server: add two or three connectors read-only, set permissions and approvals, and check it works.

By The SpanDock team

5 min read

This guide sets up the SpanDock MCP gateway with two or three connectors, read-only to start with. When you finish, every computer's AI tools will call one MCP server, your server will decide what each person may do, and every call will be recorded. Starting read-only lets your team get value from the connectors before anyone can change anything through them.

How the MCP gateway works

The MCP gateway puts one MCP server, SpanDock's, in front of your organisation's MCP connectors: knowledge bases, issue trackers, code hosts, design tools and internal APIs.

  • One MCP server on every computer. AI tools call SpanDock's MCP server on their own computer. The client forwards each call to your server over the encrypted tunnel.
  • The server decides. It checks who is calling and applies permissions, rules and guardrails before it calls the connector.
  • Credentials stay on the server. Connector tokens and sign-ins are kept in the server's credential store. They are never shown again, never logged and never sent to computers.
  • Shared sign-in or each person's own account. A connector can use one shared sign-in, or each person's own account so that the service's own permissions apply as well. SpanDock's rules still apply first, and can only narrow that access.

Prerequisites

  • A SpanDock server on the Premium or Enterprise plan, or in the 30-day trial. The gateway isn't on other plans.
  • Administrator access to the server's dashboard.
  • Computers paired to the server as clients. See Pairing clients if you haven't done this yet.
  • For each connector you plan to add, either a token for the service or an account that can approve SpanDock in a browser.

Step 1: pick your first connectors

Open MCP & connectors → MCP gateway → Catalog on the server's dashboard. The catalogue has 142 connectors; the Integrations page lists them all.

Choose two or three that your team would read from most often, such as a knowledge base and an issue tracker. Read access to those already answers many everyday questions, and it carries little risk.

Connectors come in two kinds:

  • Token connectors: paste the service's token.
  • Sign-in connectors: choose Connect and approve SpanDock once in your browser.

If your team runs its own MCP server, you can add it too, over HTTPS or as a program run on the server.

Step 2: check permissions

Every connector starts read-only for everyone. Read-only access hides write and destructive tools completely: they aren't listed and can't be called. For a first rollout, that default is usually what you want, so you may not need to change anything.

When you do need differences, each connector has a default plus overrides by person, group, tag or user type (such as Employee or Contractor). An override can:

  • turn access on or off;
  • set read-only, or read and write;
  • allow or deny tools by name;
  • require approval for some tools;
  • limit calls a minute.

Two rules decide what applies:

  1. The most specific level wins: person, then group and tag, then user type, then the connector's default.
  2. When a person's groups disagree, the most restrictive setting applies.

Step 3: decide on approvals

When you later grant write access, a connector's destructive tools (deleting, merging, closing) still need approval. A tool that needs approval doesn't run. The AI tool is told that the call is waiting, and an administrator reviews it on Approvals: who asked, which tool, and the arguments with secrets removed.

Approve once lets exactly that call through, once, within 24 hours. A denied call stays denied.

Step 4: add rules from your policies (optional)

Under Safety & compliance → Guardrails → Policies, upload or link your company policies. AI tools can then search them through the gateway.

Suggest rules reads those policies with a model you choose and proposes rules, each quoting the passage it comes from: block a tool, require approval for a tool, make a connector read-only, or redact emails, phone numbers, card numbers, secrets or your own pattern.

Nothing applies by itself. An administrator accepts, edits or dismisses each suggestion. You can also write rules by hand. A block or read-only rule can't be overridden by a permission.

Step 5: turn it on

  1. In MCP & connectors → MCP gateway, confirm the connectors you added in step 1.
  2. Check each connector's default and overrides.
  3. Turn the gateway on.
  4. On Computers, set Every computer to Recommended (SpanDock adds its MCP server to each AI tool it finds) or Enforced (it also reports, or moves aside, the tools' other MCP servers).

SpanDock configures Claude Code, Codex, Gemini CLI, Cursor, VS Code (Copilot), Windsurf and Claude Desktop. Every change to a tool's settings is backed up first, and turning the setting off undoes it exactly.

Verify it works

  1. Open Permissions → Effective permissions. Pick a person, computer or user type and check what they get for every tool, and why.
  2. On a paired computer, ask an AI tool something that needs one of your connectors, so that it calls a read tool.
  3. Open Activity on the server. The call is recorded with the connector and tool, the outcome and why, the person and computer, the duration, and how many items were redacted.

Arguments and results are recorded only if an administrator turns that on, with a notice to the people concerned.

Troubleshooting

  • A tool is missing from the AI tool's list. If the person has read-only access, write and destructive tools are hidden by design. Check Permissions → Effective permissions for the reason.
  • A call was refused. A refused call returns an error that says why, and the AI tool shows it to the person. Causes include a block rule, a block pattern in the arguments, a size limit or a rate limit.
  • A call is waiting. It needs approval. Review it on Approvals.

Next steps

  • On Enterprise, guardrails on computers also apply to gateway calls, and can keep secrets and personal data out of prompts.
  • On Enterprise, and in the trial, just-in-time access lets you make a connector's write tools nobody's standing access: people request them for a while, with a reason, and an approver you name grants them. The MCP gateway docs describe how to set it up.
  • AI tools can also use the gateway's built-in servers, including SpanDock's own read-only telemetry tools. See MCP and the assistant.
Cookie categories

Details: Cookie Policy.