> ## Documentation Index
> Fetch the complete documentation index at: https://ngrok.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# ngrok MCP Server Tools and Scopes

> Learn which ngrok MCP server tools read, change, or delete resources in your account, and the personal access token scope each tool needs.

This page lists every tool the [ngrok MCP server](https://ngrok.com/docs/mcp-server.md) exposes, what each one does, and the scopes it needs.
Use it to decide what to approve on the OAuth consent screen, or which scopes to give a [personal access token](https://ngrok.com/docs/mcp-server.md#clients-without-oauth).

## How to read the tables

The **Type** column matches the markers the server sends to your client, which [Make changes safely](https://ngrok.com/docs/mcp-server.md#make-changes-safely) explains:

* **Read-only**: The tool only reads data.
* **Changes**: The tool creates or updates something in your account, or sends something on your behalf.
* **Destructive**: The tool can delete, revoke, or break something that's in use, so your client should always ask before it runs.

The **Scopes** column lists the scopes every call of the tool needs.
If you signed in with OAuth and your token lacks one that your role allows, the server asks your client to sign in again to approve it.
If your personal access token lacks one, the tool returns a scope error.
A dash (—) means the tool needs no scope of its own.
The ngrok API still checks your role and your scopes on every call, and some arguments need an extra scope.

Some accounts see more tools than this page lists, such as tools for products in early access.

## Account and team

| Tool | Type | Scopes |
| - | - | - |
| `platform_whoami` Shows the user or service user calling, the account, the caller's role, and the plan. | Read-only | — |
| `platform_list_team` Lists users with their roles, service users, and pending invitations. | Read-only | `team:read`, `service-users:read` |
| `platform_invite_team_member` Sends an email invitation to join the account. | Changes | `ngrok:invitations:write` |
| `platform_list_api_keys` Lists API keys with their owner and status, never the key itself. | Read-only | `api-keys:read` |
| `platform_list_authtokens` Lists authtokens with their owner, ACL, and status, never the token itself. | Read-only | `credentials:read` |
| `platform_list_ssh_keys` Lists SSH public keys registered for the agentless SSH tunnel. | Read-only | `ssh-credentials:read` |
| `platform_list_access_tokens` Lists personal access tokens and when each expires or was revoked, never the token itself. | Read-only | `credentials:read` |
| `platform_get_usage` Shows what the account consumed in a window, next to what the plan includes. | Read-only | `ngrok:usage:read` |
| `platform_get_endpoint_usage` Ranks the endpoints that drove usage in a window. | Read-only | `ngrok:usage:read` |
| `platform_get_limits` Shows each plan limit, current usage, and when a limit last blocked the account. | Read-only | `ngrok:limits:read` |

To revoke an authtoken, API key, or SSH key, or to cancel an invitation, the AI agent uses [`delete_resource`](#get-and-delete-any-resource).

## Endpoints, agents, and traffic

| Tool | Type | Scopes |
| - | - | - |
| `gw_list_endpoints` Lists endpoints with their URL, type, binding, upstream, and whether they have a Traffic Policy. | Read-only | `endpoints:read` |
| `gw_create_cloud_endpoint` Creates a Cloud Endpoint. | Changes | `endpoints:write` |
| `gw_update_cloud_endpoint` Updates a Cloud Endpoint's URL, bindings, pooling, or Traffic Policy. | Changes | `endpoints:read`, `endpoints:write` |
| `gw_validate_traffic_policy` Checks a Traffic Policy document for errors without saving it. | Read-only | — |
| `gw_list_agents` Lists connected ngrok agents with their version, deprecation deadline, OS, and region. | Read-only | `tunnel-sessions:read` |
| `gw_list_kubernetes_operators` Lists Kubernetes Operators registered to the account. | Read-only | `kubernetes-operators:read` |
| `gw_query_traffic_inspector` Finds HTTP requests in Traffic Inspector, or counts them by field or by time period. | Read-only | `events:read` |
| `gw_get_traffic_inspector_request` Shows one request in full, including headers and bodies when full capture is on. | Read-only | `events:read` |
| `gw_query_connections` Finds open and closed TCP and TLS connections, or counts them by field or by time period. | Read-only | `events:read` |
| `gw_get_connection` Shows one connection in full, including timing, bytes, TLS details, and failure. | Read-only | `events:read` |
| `gw_get_connection_progress` Shows the bytes a TCP or TLS connection moved over time. | Read-only | `events:read` |
| `gw_list_app_users` Lists people who signed in to your endpoints through an OAuth, OIDC, or SAML action. | Read-only | `app-users:read` |

## Domains, addresses, and certificates

| Tool | Type | Scopes |
| - | - | - |
| `gw_list_domains` Lists reserved domains with their CNAME target and certificate status. | Read-only | `domains:read` |
| `gw_reserve_domain` Reserves a domain. | Changes | `domains:write` |
| `gw_update_domain` Updates a reserved domain. | Changes | `domains:write` |
| `gw_list_tcp_addresses` Lists reserved TCP addresses. | Read-only | `tcp-addrs:read` |
| `gw_reserve_tcp_address` Reserves a TCP address. | Changes | `tcp-addrs:write` |
| `gw_update_tcp_address` Updates a reserved TCP address's description or metadata. | Changes | `tcp-addrs:write` |
| `gw_list_tls_certificates` Lists uploaded TLS certificates. | Read-only | `tls-certificates:read` |
| `gw_list_certificate_authorities` Lists certificate authorities used for mutual TLS. | Read-only | `certificate-authorities:read` |

## IP policies

| Tool | Type | Scopes |
| - | - | - |
| `gw_list_ip_policies` Lists IP policies. | Read-only | `ip-policies:read` |
| `gw_list_ip_policy_rules` Lists IP policy rules across every policy. | Read-only | `ip-policies:read` |
| `gw_create_ip_policy` Creates an IP policy. | Changes | `ip-policies:write` |
| `gw_update_ip_policy` Updates an IP policy's description or metadata. | Changes | `ip-policies:read`, `ip-policies:write` |
| `gw_create_ip_policy_rule` Adds a rule that allows or denies a CIDR. | Changes | `ip-policies:write` |
| `gw_update_ip_policy_rule` Updates an IP policy rule. | Changes | `ip-policies:read`, `ip-policies:write` |

## Vaults and secrets

Secret values never go through the chat.
`gw_create_secret` and `gw_update_secret` send you a dashboard link where you type the value, and the dashboard checks your permissions when you save it.

| Tool | Type | Scopes |
| - | - | - |
| `gw_list_vaults` Lists vaults. | Read-only | `vaults:read` |
| `gw_list_secrets` Lists secret names and metadata, never values. | Read-only | `vaults:read` |
| `gw_create_vault` Creates a vault. | Changes | `vaults:write` |
| `gw_update_vault` Updates a vault's name, description, or metadata. | Changes | `vaults:write` |
| `gw_create_secret` Sends you a dashboard link to create a secret and type its value. | Changes | `vaults:read` |
| `gw_update_secret` Updates a secret's name, description, or metadata, or sends you a link to set a new value. | Changes | — |

## Log exports

| Tool | Type | Scopes |
| - | - | - |
| `gw_list_event_subscriptions` Lists event subscriptions, which choose what the account exports as logs. | Read-only | `events:read` |
| `gw_list_event_destinations` Lists event destinations and their recent delivery errors, never their credentials. | Read-only | `events:read` |

## Get and delete any resource

Two tools work across resource types.
The ID prefix picks the type, and the ngrok API checks the scope for that type.

| Tool | Type | Scopes |
| - | - | - |
| `get_resource` Reads one object by its ID or by a dashboard URL, or an endpoint by its URL. | Read-only | — |
| `delete_resource` Deletes, revokes, or releases one object by its ID. | Destructive | — |

`delete_resource` can:

* Delete a Cloud Endpoint, but not an Agent Endpoint or an AI Gateway endpoint.
* Release a reserved domain or TCP address.
* Delete an IP policy with all its rules, or one IP policy rule.
* Delete an empty vault, or a secret.
* Revoke an authtoken, API key, or SSH public key.
* Cancel a pending team invitation.
* Revoke an AI Gateway access key, or delete an AI Gateway access config, custom provider, provider key, virtual model, or firewall policy.

## Documentation

These tools read the public ngrok documentation, so they need no scope.

| Tool | Type | Scopes |
| - | - | - |
| `docs_search` Searches the ngrok documentation. | Read-only | — |
| `docs_get_page` Reads one documentation page. | Read-only | — |
| `docs_get_error` Looks up an `ERR_NGROK` error code. | Read-only | — |

## AI Gateway

For the usage, request, capture, and model tools, such as `aigw_list_usage` and `aigw_get_model`, see [AI Gateway tools](https://ngrok.com/docs/mcp-server/ai-gateway.md#tools-and-scopes).

The tools in this section read and change your AI Gateway setup.
A change applies to live traffic right away.
Access key tokens and provider API keys never go through the chat: `aigw_create_access_key`, `aigw_create_provider_key`, and `aigw_rotate_provider_key` send you an AI dashboard link where you finish the step, so they need only a read scope on the token, and the dashboard checks your permissions when you save.

| Tool | Type | Scopes |
| - | - | - |
| `aigw_list_access_configs` Lists access configs, which limit the providers and models a key may use. | Read-only | `ai-gateway:access-keys:read` |
| `aigw_get_access_config` Shows one access config. | Read-only | `ai-gateway:access-keys:read` |
| `aigw_create_access_config` Creates an access config. | Changes | `ai-gateway:access-keys:write` |
| `aigw_update_access_config` Updates an access config, which changes what every key that uses it may do. | Destructive | `ai-gateway:access-keys:write` |
| `aigw_list_access_keys` Lists access keys, never their tokens. | Read-only | `ai-gateway:access-keys:read` |
| `aigw_get_access_key` Shows one access key, never its token. | Read-only | `ai-gateway:access-keys:read` |
| `aigw_create_access_key` Sends you an AI dashboard link to create an access key. | Changes | `ai-gateway:access-keys:read` |
| `aigw_update_access_key` Updates an access key. | Changes | `ai-gateway:access-keys:write` |
| `aigw_list_providers` Lists custom providers. | Read-only | `ai-gateway:providers:read` |
| `aigw_get_provider` Shows one custom provider. | Read-only | `ai-gateway:providers:read` |
| `aigw_create_provider` Creates a custom provider. | Changes | `ai-gateway:providers:write` |
| `aigw_update_provider` Updates a custom provider. | Destructive | `ai-gateway:providers:write` |
| `aigw_list_provider_keys` Lists provider keys, never their values. | Read-only | `ai-gateway:providers:read` |
| `aigw_get_provider_key` Shows one provider key, never its value. | Read-only | `ai-gateway:providers:read` |
| `aigw_create_provider_key` Sends you an AI dashboard link to add a provider key. | Changes | `ai-gateway:providers:read` |
| `aigw_rotate_provider_key` Sends you an AI dashboard link to replace a provider key's value. | Changes | `ai-gateway:providers:read` |
| `aigw_update_provider_key` Updates a provider key. | Changes | `ai-gateway:providers:write` |
| `aigw_list_virtual_models` Lists virtual models. | Read-only | `ai-gateway:virtual-models:read` |
| `aigw_get_virtual_model` Shows one virtual model and its target models. | Read-only | `ai-gateway:virtual-models:read` |
| `aigw_create_virtual_model` Creates a virtual model. | Changes | `ai-gateway:virtual-models:write` |
| `aigw_update_virtual_model` Updates a virtual model. | Destructive | `ai-gateway:virtual-models:write` |
| `aigw_list_firewall_policies` Lists firewall policies. | Read-only | `ai-gateway:settings:read` |
| `aigw_get_firewall_policy` Shows one firewall policy. | Read-only | `ai-gateway:settings:read` |
| `aigw_create_firewall_policy` Creates a firewall policy. | Changes | `ai-gateway:settings:write` |
| `aigw_update_firewall_policy` Updates a firewall policy. | Destructive | `ai-gateway:settings:write` |
| `aigw_get_gateway_config` Shows the account-wide AI Gateway settings. | Read-only | `ai-gateway:settings:read` |
| `aigw_update_gateway_config` Updates the account-wide AI Gateway settings. | Destructive | `ai-gateway:settings:write` |
| `aigw_add_credits` Gives you the AI dashboard billing page. It buys nothing. | Read-only | — |

## Feedback

| Tool | Type | Scopes |
| - | - | - |
| `send_feedback` Reports a missing, confusing, or broken ngrok MCP tool to ngrok. | Changes | — |

## Tools by scope

When you create a personal access token, give it the scopes for the tools you want your AI agent to use.
A tool runs only when the token has every scope the tool lists.

| Scope | Tools it unlocks |
| - | - |
| `endpoints:read` | `gw_list_endpoints`, `gw_update_cloud_endpoint` (with `endpoints:write`) |
| `endpoints:write` | `gw_create_cloud_endpoint`, `gw_update_cloud_endpoint` (with `endpoints:read`) |
| `events:read` | `gw_query_traffic_inspector`, `gw_get_traffic_inspector_request`, `gw_query_connections`, `gw_get_connection`, `gw_get_connection_progress`, `gw_list_event_subscriptions`, `gw_list_event_destinations` |
| `tunnel-sessions:read` | `gw_list_agents` |
| `kubernetes-operators:read` | `gw_list_kubernetes_operators` |
| `app-users:read` | `gw_list_app_users` |
| `domains:read` | `gw_list_domains` |
| `domains:write` | `gw_reserve_domain`, `gw_update_domain` |
| `tcp-addrs:read` | `gw_list_tcp_addresses` |
| `tcp-addrs:write` | `gw_reserve_tcp_address`, `gw_update_tcp_address` |
| `tls-certificates:read` | `gw_list_tls_certificates` |
| `certificate-authorities:read` | `gw_list_certificate_authorities` |
| `ip-policies:read` | `gw_list_ip_policies`, `gw_list_ip_policy_rules`, `gw_update_ip_policy` and `gw_update_ip_policy_rule` (with `ip-policies:write`) |
| `ip-policies:write` | `gw_create_ip_policy`, `gw_create_ip_policy_rule`, `gw_update_ip_policy` and `gw_update_ip_policy_rule` (with `ip-policies:read`) |
| `vaults:read` | `gw_list_vaults`, `gw_list_secrets`, `gw_create_secret` |
| `vaults:write` | `gw_create_vault`, `gw_update_vault` |
| `team:read` | `platform_list_team` (with `service-users:read`) |
| `service-users:read` | `platform_list_team` (with `team:read`) |
| `ngrok:invitations:write` | `platform_invite_team_member` |
| `api-keys:read` | `platform_list_api_keys` |
| `credentials:read` | `platform_list_authtokens`, `platform_list_access_tokens` |
| `ssh-credentials:read` | `platform_list_ssh_keys` |
| `ngrok:usage:read` | `platform_get_usage`, `platform_get_endpoint_usage` |
| `ngrok:limits:read` | `platform_get_limits` |
| `ai-gateway:usage:read` | `aigw_usage_overview`, `aigw_list_usage`, and with `ai-gateway:captures:read`: `aigw_get_request`, `aigw_get_request_attempts`, `aigw_get_capture` |
| `ai-gateway:captures:read` | With `ai-gateway:usage:read`: `aigw_get_request`, `aigw_get_request_attempts`, `aigw_get_capture` |
| `ai-gateway:providers:read` | `aigw_list_models`, `aigw_get_model`, `aigw_list_providers`, `aigw_get_provider`, `aigw_list_provider_keys`, `aigw_get_provider_key`, `aigw_create_provider_key`, `aigw_rotate_provider_key` |
| `ai-gateway:providers:write` | `aigw_create_provider`, `aigw_update_provider`, `aigw_update_provider_key` |
| `ai-gateway:access-keys:read` | `aigw_list_access_configs`, `aigw_get_access_config`, `aigw_list_access_keys`, `aigw_get_access_key`, `aigw_create_access_key` |
| `ai-gateway:access-keys:write` | `aigw_create_access_config`, `aigw_update_access_config`, `aigw_update_access_key` |
| `ai-gateway:virtual-models:read` | `aigw_list_virtual_models`, `aigw_get_virtual_model` |
| `ai-gateway:virtual-models:write` | `aigw_create_virtual_model`, `aigw_update_virtual_model` |
| `ai-gateway:settings:read` | `aigw_list_firewall_policies`, `aigw_get_firewall_policy`, `aigw_get_gateway_config` |
| `ai-gateway:settings:write` | `aigw_create_firewall_policy`, `aigw_update_firewall_policy`, `aigw_update_gateway_config` |

Some tools need no scope of their own, such as `platform_whoami`, `gw_validate_traffic_policy`, and the docs tools.
`get_resource` and `delete_resource` need none either, but the ngrok API checks the token's scopes for the type of object they read or delete.

## Related resources

* [ngrok MCP server](https://ngrok.com/docs/mcp-server.md): Connect your AI agent and sign in.
* [AI Gateway tools](https://ngrok.com/docs/mcp-server/ai-gateway.md): Ask about AI Gateway models, costs, and requests.
* [Example prompts](https://ngrok.com/docs/mcp-server/prompts.md): Prompts you can copy, grouped by job.
* [Role-based access control](https://ngrok.com/docs/iam/rbac.md): What your role allows the AI agent to do.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.