read_capability
Execute read-only Prisma SD-WAN API actions from the registry by ID for direct data access when semantic tools lack coverage.
Instructions
Execute one exact read-only registry capability as an expert escape hatch.
Prefer the semantic tools first. Use list_capabilities to discover an
action_id. Parameters are validated against the registry and every response
passes through central recursive secret redaction and response-size limits.
Curated actions that still need live verification are blocked here by default; set MCP_ALLOW_UNVERIFIED_COMPAT=true only after completing the validation checklist shipped with this project.
Args:
action_id: Exact action identifier from list_capabilities, e.g.
"sites_devices.sites". Not a free-text search term.
path_parameters: One key per required/optional path parameter that
action's list_capabilities entry lists under
path_parameters (e.g. {"site_id": "..."}). Omit or use
{} for actions with none. Unknown keys are rejected.
body: JSON object matching that action's body_schema. Only
meaningful for POST actions — passing any non-empty body to
a GET action is rejected, as is any field the action's
schema does not declare (the error names the valid fields).
Two traps on the standard query body: retrieved_fields is
refused — the controller answers it with the full record
and a fabricated value in every unrequested field, so the reply
looks right and is wrong; and a limit inside the body is
not this tool's limit — it truncates at the controller and
reports the truncated total with no cursor, stranding the rest.
Page with the limit/cursor arguments below instead.
cursor: Opaque pagination token copied from a previous response's
next_cursor. Only applies when the result is a list. A cursor
is bound to the tool that issued it — one from another tool is
rejected rather than applied to an unrelated list.
limit: Max items to return in this page when the result is a list.
Omit to use the server default page size (50; max 200). This is the tool's own
paging, applied after the response arrives — prefer it over a
limit field in body.
detail: "compact" (default) lets a large response drop fields that
neither identify a record nor report its condition, so more records
fit in one reply; whatever it held back is listed in
omitted_fields. Pass "full" to get every field of every
record instead, at the cost of fewer records per page. Small
responses are unaffected either way.
Tool call template
Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.
{
"id": 1,
"jsonrpc": "2.0",
"method": "tools/call",
"params": {
"arguments": {
"action_id": "<action_id>"
},
"name": "read_capability"
}
}Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| body | No | JSON object matching that action's ``body_schema``. Only meaningful for ``POST`` actions — passing any non-empty body to a ``GET`` action is rejected, as is any field the action's schema does not declare (the error names the valid fields). Two traps on the standard query body: ``retrieved_fields`` is **refused** — the controller answers it with the full record and a fabricated value in every unrequested field, so the reply looks right and is wrong; and a ``limit`` *inside* the body is not this tool's `limit` — it truncates at the controller and reports the truncated total with no cursor, stranding the rest. Page with the `limit`/`cursor` arguments below instead. | |
| limit | No | Max items to return in this page when the result is a list. Omit to use the server default page size (50; max 200). This is the tool's own paging, applied after the response arrives — prefer it over a ``limit`` field in `body`. | |
| cursor | No | Opaque pagination token copied from a previous response's `next_cursor`. Only applies when the result is a list. A cursor is bound to the tool that issued it — one from another tool is rejected rather than applied to an unrelated list. | |
| detail | No | ``"compact"`` (default) lets a large response drop fields that neither identify a record nor report its condition, so more records fit in one reply; whatever it held back is listed in ``omitted_fields``. Pass ``"full"`` to get every field of every record instead, at the cost of fewer records per page. Small responses are unaffected either way. | |
| action_id | Yes | Exact action identifier from `list_capabilities`, e.g. ``"sites_devices.sites"``. Not a free-text search term. | |
| path_parameters | No | One key per required/optional path parameter that action's `list_capabilities` entry lists under ``path_parameters`` (e.g. ``{"site_id": "..."}``). Omit or use ``{}`` for actions with none. Unknown keys are rejected. |
Output Schema
| Name | Required | Description | Default |
|---|---|---|---|
| result | Yes |