Claude GoPlus AgentGuard connectorWhat Claude can do with GoPlus AgentGuard on your machine.
The Claude GoPlus AgentGuard connector exposes 7 tools. 5 read or evaluate without changing anything, 2 change your local trust registry, 0 are undocumented. Here: what Claude can check before you trust a skill, and what this connector does not block.
Verified Trustpilot reviews · AI, automation & growth agency
What the Claude GoPlus AgentGuard connector changes for agent security
Before you hand an agent a new skill, or let it run a command or sign a transaction, you want a second opinion. This connector, a desktop extension that runs on your own machine, lets Claude ask GoPlus AgentGuard for one: scan a skill folder for risky content, check a trust registry, weigh a proposed action against policy, or simulate a Web3 transaction before anyone signs it.
A skill checked before it is trusted. A developer downloads a community skill and asks Claude whether it is safe. skill_scanner_scan scans the skill directory for malicious or risky content, and registry_lookup shows whether that skill is already in the trust registry.
A verdict on a risky step. An agent wants to run something that looks destructive. action_scanner_decide evaluates the proposed action against policy and returns allow or deny, which Claude can explain in plain words.
A transaction tested, not signed. A crypto user is about to approve a token swap. action_scanner_simulate_web3 simulates it to surface risks before signing.
What it does not do: GoPlus itself says live action blocking goes through its own hooks, and that this connector is best for scans that an agent requests, registry operations and Web3 simulation. In other words, calling a tool from the conversation does not stand guard over everything else running on your machine. The registry is local, so an attestation stays on that computer. And nothing runs on its own schedule: Claude calls these tools while it answers you. If you need recurring checks, look at an automation tool, a different job described on the Integrations hub.
The vocabulary in one minute
Five words you will see while adding AgentGuard.
- Connector
- The link you set up once between Claude and another service or program, so Claude can use it while it answers you.
- Tool
- One named action a connector gives Claude. Claude decides which ones to call; the directory lists each one under its exact name.
- Authorization
- The screen where you grant the access Claude will use, when the service shows one. Granted once per person, and it can be withdrawn.
- Approval
- The confirmation Claude waits for before it goes ahead with something that changes data, shown in the conversation at that moment.
- MCP
- The shared standard connectors are built on: it lets an assistant like Claude talk to an outside service or a program on your computer.
Add GoPlus AgentGuard to Claude in three steps
- 01
Open Claude Desktop
AgentGuard is a desktop extension, so it lives in Claude Desktop. Open its settings, go to Customize then Connectors, and find GoPlus AgentGuard. On a Team or Enterprise workspace, an Owner or Primary Owner has to enable it before members can use it.
- 02
Install and enable it
Start the setup from its row. The publisher documents no extra step for Claude Desktop beyond its requirements, so what you see on screen comes from Claude Desktop itself. If the setup breaks later, Disconnect it and add it again.
- 03
Read what you allow
If a permission or authorization screen appears, read it before you accept. It belongs to the service, not to Claude, and it sets how far the access goes.
The 7 tools, sorted by what they do
GoPlus AgentGuard gives Claude 7 tools: 5 that read your account, 2 that change something in it.
Five tools inspect or evaluate, two change the local trust registry. Names stay exactly as Claude shows them.
- 5 read
- 2 write
What Claude checks (5)
5 toolsFive tools that scan, look up, list, evaluate or simulate without changing anything.
skill_scanner_scan
Inspects an agent skill directory for malicious or risky content and reports what it finds, so Claude can tell you whether a skill looks safe before you rely on it.
registry_lookup
Checks whether a given skill or package is recorded in the trust registry, and what the registry says about it. The lookup only reads; the registry stays as it was.
registry_list
Returns every entry in the local trust registry, giving Claude the full picture of what has been marked as trusted on this machine. Reading the list leaves every entry as it was.
action_scanner_decide
Evaluates a proposed runtime action against policy and returns allow or deny. Claude reads the verdict back to you instead of guessing whether a step is dangerous.
action_scanner_simulate_web3
Runs a Web3 transaction in simulation to surface its risks before signing, so nothing reaches the chain while Claude explains the outcome.
What Claude changes (2)
2 toolsTwo tools that add or remove trust in the local registry. No source describes a confirmation specific to either, so the general rule below applies.
registry_attest
Approval: see the ruleMarks a skill or package as trusted in the local registry.
registry_revoke
Approval: see the ruleWithdraws a previous attestation from the registry, so a skill that was trusted no longer is.
What Claude asks before it acts
By default, Claude stops and asks for your go-ahead before any action it would carry out for you. That covers attesting and revoking here, since no source sets a different rule for those two tools.
On Team and Enterprise, workspace owners decide whether members may let some actions through without a fresh confirmation, and they can limit what a connector may do for the whole organization; nobody overrides that from a personal account. Claude works with the rights of the person using it and nothing more. Here the extension runs locally, and the registry it changes is the one on that machine, not a shared one.
Which plans it is available on
None of the 819 sheets in the official directory shows plan availability. Connector by connector, that answer is published nowhere: a real gap in the catalogue.
The general rule is published: remote connectors are open to all users on Claude, Cowork, Claude Desktop and mobile, while desktop extensions like this one install in Claude Desktop. On Team and Enterprise, an Owner or Primary Owner enables a connector for the organization before members use it. For the current status, check the GoPlus AgentGuard sheet in the official directory.
Where the Claude GoPlus AgentGuard connector stops
A connector is not an automation. Claude calls these tools while it answers you; no scan starts on its own when a new skill lands on your disk.
Only GoPlus documents this connector, in its own repository; Claude's help center has no article on it. A list of tools is an observed floor, not a promise, since an administrator can open actions that appear on no public sheet, and the partner badge is not a security audit: Anthropic says on every sheet that it neither chooses the tools a publisher exposes nor guarantees how they behave. Connect a security tool only if you trust its publisher.
Need help connecting GoPlus AgentGuard to Claude?
A person reads every message.