---
title: "Read one connected MCP server's dial state"
method: GET
path: "/v1/mcp-servers/{account_id}"
tags: ["mcp-servers"]
---

# Read one connected MCP server's dial state

`GET /v1/mcp-servers/{account_id}`

## Request

- Base URL: `https://api.patricia.app` (the document declares no server; this is the origin it was published from)
- URL: `https://api.patricia.app/v1/mcp-servers/{account_id}`
- Auth: HTTP bearer

The address, auth mode, header name and refresh state, plus the last catalog reading — what the server offered and how much of it Patricia could accept. NEVER the token: it is stored sealed and is not readable by any route.

## Path parameters

- `account_id` string, uuid, required

## Headers

- `X-Tenant-Id` string, nullable

## Response `200`

Successful Response

- McpServerResponse: What the dashboard renders on the card.
  - `integration_account_id` string, required
  - `external_account_id` string, required
  - `base_url` string, required
  - `auth_mode` string, required
  - `header_name` string, nullable
  - `negotiated_protocol_version` string, nullable
  - `connection_generation` integer, required
  - `refreshing` boolean, required
  - `last_error_class` string, nullable
  - `needs_reauth` boolean
  - `connected_by` string, nullable
  - `catalog` McpCatalogOutcomeResponse: The last catalog reading recorded for a server — what it offered and how much Patricia took. This exists because a refusal was previously invisible: a tool whose schema Patricia declines yields no spec, so it is simply ABSENT from the listing, indistinguishable from one the server withdrew. The card could show a server as connected and healthy while half its catalog had been refused, and no surface said so. Every field describes THE READING, never the account's current rows. On a refused reduction the enabled rows are the PREVIOUS catalog, so the two answer different questions and must not be mixed in one sentence.
    - `complete` boolean, required
    - `accepted_unique` integer, required
    - `offered_unique` integer, required
    - `rejected` integer, required
    - `dropped_pre_schema` integer, required
    - `reasons` object[], required
    - `observed_at` string, date-time, required
  - `visibility` string, required
  - `is_mine` boolean
  - `can_manage` boolean
  - `can_disconnect` boolean
  - `can_reconsent` boolean
  - `can_share_to_team` boolean
  - `can_make_private` boolean
  - `share_to_team_unavailable_reason` 'already_shared' | 'managed_account' | 'scope_sealed_at_connect' | 'missing_edit_permission' | 'not_your_account' | 'mcp_share_needs_reconnect': Why ``can_share_to_team`` is False for this caller. Null when it is True. The value names the first term of that answer that fails, in the order ``account_capabilities`` states: - ``already_shared``: the connection is shared with the whole workspace already. - ``managed_account``: Patricia manages the connection. The visibility write refuses a managed connection for every caller. - ``scope_sealed_at_connect``: the connection's credential is sealed to the sharing it was connected with (``core.scope_sealed_credentials``), and both visibility writes refuse to move it in either direction. Its refusal message carries the remedy, ``core.scope_sealed_credentials.SCOPE_SEALED_REMEDY``. - ``missing_edit_permission``: the caller does not hold ``integrations:edit``. - ``not_your_account``: the caller is not the person who connected this private connection. - ``mcp_share_needs_reconnect``: the MCP server's saved sign-in would stop its tools working once the server is shared, so the write refuses with the ``MCP_SHARE_NEEDS_RECONNECT`` error code. Reconnecting the server is the remedy.
  - `make_private_unavailable_reason` 'already_private' | 'managed_account' | 'scope_sealed_at_connect' | 'missing_edit_permission' | 'owner_unknown' | 'oauth_owner_mismatch': Why ``can_make_private`` is False for this caller. Null when it is True. The value names the first term of that answer that fails, in the order ``account_capabilities`` states: - ``already_private``: the connection is private already. - ``managed_account``: Patricia manages the connection. The visibility write refuses a managed connection for every caller. - ``scope_sealed_at_connect``: the connection's credential is sealed to the sharing it was connected with (``core.scope_sealed_credentials``), and both visibility writes refuse to move it in either direction. Its refusal message carries the remedy, ``core.scope_sealed_credentials.SCOPE_SEALED_REMEDY``. - ``missing_edit_permission``: the caller does not hold ``integrations:edit``. - ``owner_unknown``: the connection records nobody it could become private to. An MCP server needs a ``connected_by_user_id``; another connection needs an email or actor. - ``oauth_owner_mismatch``: the MCP server signs in with OAuth, and its saved sign-in is not the recorded connector's: there is none, it names another user or no membership, or its membership is not the one the connector's owner contract resolves. The narrowing's grant check refuses it with the ``MCP_OAUTH_OWNER_MISMATCH`` error code. Disconnecting the server and having that person connect it again as private is the remedy.

## Other responses

- `422`: Validation Error

---

[API](https://skmtc.dev/patricia/apis/patricia-api.md) · [All operations](https://skmtc.dev/patricia/apis/patricia-api/llms.txt) · [OpenAPI document](https://skmtc.dev/patricia/apis/patricia-api/revisions/0dfb44495ea1?raw)