teamgrid_roles_list is a read-only TeamGrid MCP tool. It is advertised in: full, admin-write.
List workspace roles
Arguments at a glance
The table describes full. When a profile variant is shown below, use that variant’s exact schema.
| Argument | Presence | Type | Meaning |
|---|---|---|---|
cursor |
Optional | string | See the exact schema below for values and constraints. |
limit |
Optional | integer | See the exact schema below for values and constraints. |
Input schema
Stable release: this is the exact JSON Schema advertised by @teamgrid/mcp-server@1.2.2:
{
"type": "object",
"properties": {
"cursor": {
"maxLength": 1024,
"type": "string"
},
"limit": {
"default": 50,
"maximum": 100,
"minimum": 1,
"type": "integer"
}
},
"required": [],
"additionalProperties": false
}The schema above is for full. Properties not in the selected profile schema are rejected.
Scope and API operation
Required scopes (all): roles:read.
listRoles—GET /roles
The credential must also satisfy normal workspace authorization and any service-account resource grants. Selecting an MCP tool profile never adds scopes to a credential.
Read behavior
This tool does not change business state.
Output and limits
The bounded API result is returned as MCP structured content and equivalent JSON text. This is a cursor-paginated tool. limit accepts 1–100. When meta.page.nextCursor is not null, pass that opaque value as cursor to request the next page. Do not construct, edit, or decode cursors. The serialized result may not exceed
256 KiB.
Download the exact MCP input and output contract, including profile variants, scopes and safety annotations. Its outputSchema describes the advertised MCP envelope and local $defs; it includes MCP projection metadata in addition to the underlying API schema.
The linked API operation describes the business resource and its field semantics. Write tools preserve their declared revision/idempotency contract and require current permissions. Accepted jobs provide status/resume information; uncertain writes must not be replayed blindly.
Security classification
governance-data: The response can contain access configuration, activity records or automation metadata. Current role and domain permissions still apply.
The exact safety annotations are {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":false}. The host and model can still retain tool
arguments and results in prompts, logs, or transcripts; use a dedicated least-privilege credential.
Example prompt
List the first 20 TeamGrid roles. Do not request another page; tell me whether another cursor is available.
The prompt is illustrative. Inspect the proposed tool arguments before approving access to personal, commercial, conversation, or security-configuration data.
Common failures
| Condition | Observable behavior and recovery |
|---|---|
The selected tool profile or allow/deny filter excludes teamgrid_roles_list. |
The tool is not advertised to the host. Select the narrowest profile that contains it and restart the host. |
| An argument violates this tool’s input schema: required field, type, enum, pattern, length, or range. | MCP input validation rejects the call before an API request is made. |
The credential lacks roles:read or cannot access the requested resource. |
The tool preserves a safe API error code such as insufficient_scope, with redacted detail and available status/request metadata. |
| A read exceeds the 256 KiB tool result limit. | The tool returns result_too_large. Request a smaller page or narrower filters. |
| An unknown input property is supplied. | The strict input schema rejects the call before an API request is made. |
Authentication failures that prevent the MCP process from starting are covered separately in MCP troubleshooting.