# API changes: November 17, 2026

[← Changelog](/docs/changelog) · [November 2026](/docs/changelog/2026-11)

### FIX-1117-1: retries with the same Idempotency-Key no longer create a second machine after the server's key changed

**Before**

Rebinding a server to another API key cleared its `Idempotency-Key` value. A retry of the original request with the same header no longer found the created machine and created one more: the retry was billed as a new creation. The same happened when a server was handed over to another user and when a key was recovered.

**After**

The `Idempotency-Key` value stays on the server through a rebind, a handover and a key recovery. A retry with the same header under the key that now manages the server returns that machine with the `Idempotent-Replayed: true` response header; the successful response remains HTTP 201, as it was at creation. The key's scope does not change — it is still one API key: a retry under the previous key has no link to the creation.

**Impact on integrators**

Nothing to change: existing calls keep working as before, and a retry that used to create a second machine now returns the machine already created.

### NEW-1117-2: Galaxy disk usage report

`GET /v1/infra/servers/:id/disk-usage` returns a saved galaxy disk report: application data, other data, images, build cache, writable container layers and container logs. Pass the host ID or your application's ID. The host key sees all live applications. An application key sees only applications managed by that same key, with other data aggregated without names. The request does not wake the host or start cleanup. If no successful snapshot exists, or the server is standalone, the response contains `data: null`.

The `galaxy-reference` host row in the server list and detail response now includes `diskTotalMb`, `diskFreeMb`, `diskState` and `diskProbedAt`. `access.allowedEndpoints` includes the new report read operation. See [Galaxy disk usage](/docs/infra/servers/disk-usage).

### FIX-1117-3: a Company connection stays active when the holder loses administrator rights

**Before**

A Company connection moved to `NEEDS_ADMIN` when the holder, the employee who approved it as an administrator, lost administrator rights in Bitrix24: on the scheduled re-check or on a role change event. While the holder check had not passed, [GET /v1/connect/company/status](/docs/partner-connect) showed `reason_code` `ADMIN_CHECK_REQUIRED` with the `AUTHORIZE_ADMIN` action, and working Bitrix24 calls made with the connection secret received HTTP 403 `CONNECTION_ADMIN_REQUIRED`.

**After**

Administrator rights are needed only at the moment of consent. After the rights are removed, the connection stays `ACTIVE` and works with the holder's permissions as a regular employee. The connection moves to `NEEDS_ADMIN` when the holder departs: they are dismissed, deactivated or deleted, become a guest, or can no longer authenticate using their own key and a repeat check at least 30 seconds later confirms it. While the holder check has not passed, the status shows `reason_code` `ADMIN_CHECK_REQUIRED` with the `WAIT` action, and working calls receive HTTP 503 `ADMIN_CHECK_REQUIRED` with a `Retry-After` header: retry the call later. The HTTP 200 response of a successful call and of the status poll is unchanged, and `NEEDS_ADMIN` still returns `CONNECTION_ADMIN_REQUIRED` and `AUTHORIZE_ADMIN`.

**Impact on integrators**

No action is required. A product that follows the status `action` field waits for the re-check instead of asking for an administrator. On HTTP 503 `ADMIN_CHECK_REQUIRED`, retry the call no earlier than the `Retry-After` value.

### FIX-1117-4: look up a lead by external identifiers

**Before**

`GET /v1/leads` with `filter[originId]`, including when combined with `filter[originatorId]`, returned `400 UNKNOWN_FILTER_FIELD`.

**After**

Both string filters are accepted together for an exact lead lookup by the external system and record identifiers. Leading zeros are preserved, and `_` and `%` are literal characters, including with `$in` and plain arrays of values. The `$ne`, `!` and `!=` operators use exact inequality instead of pattern matching. The `originId` response value is preserved, including an empty string.

### NEW-1117-5: Chat list actions and startup configuration

[Pin a chat](/docs/chats/management/pin) and [unpin a chat](/docs/chats/management/unpin) accept an optional `folderId` in the JSON body to select the pin scope. Calls without the field keep the default scope.

[POST /v1/chats/:dialogId/hide](/docs/chats/management/hide) hides a dialog from the user's list while preserving membership and messages. [GET /v1/chats/configuration](/docs/chats/discovery/configuration) returns, next to the counters, folders and anchors, the known list features, minimum roles for actions, the pin limit of a folder, the platform admission of folder actions, and the current user ID and known type without contact details. Missing facts remain unknown.

### NEW-1117-6: CoPilot commands in chats

`POST /v1/chats/:dialogId/messages` accepts a separate `command` body with `kind: "copilot"` and `version: 1`: draft and chat creation, manual chat title changes, model and role selection, stopping and regenerating answers, ratings, suggestions, form actions and chunked dictation transfer. A command body cannot be combined with message fields or query parameters. The platform does not automatically replay a command after an unknown outcome.

`GET /v1/users/me?include=chatConfiguration` adds chat settings, CoPilot capabilities, permissions and the CoPilot bot record. Requires both `user` and `im` scopes. Without `include`, the profile field set is unchanged.

`POST /v1/chats/:chatId/files/uploads/publish` supports CoPilot chats and an optional `copilot` object with message parameters.

### NEW-1117-7: Choose whether to show an app in Bitrix24

[POST /v1/infra/servers/:id/deploy](/docs/infra/deploy/deploy) accepts the optional `showInBitrix24` parameter. `false` deploys the app without a catalog card or chat, while `true` allows them to be created. The parameter is supported in JSON and multipart, for standalone virtual machines and Galaxy apps, and in [POST /v1/infra/servers](/docs/infra/servers/create) when deploying a Galaxy app during creation.

The initial choice is saved. A subsequent deploy with this parameter preserves the choice and adds a `SHOW_IN_BITRIX24_ALREADY_CHOSEN` warning to `warnings[]`.

[POST /v1/infra/servers/:id/b24-catalog/publish](/docs/infra/servers/b24-catalog-publish) saves the choice `YES`. The new [POST /v1/infra/servers/:id/b24-catalog/hide](/docs/infra/servers/b24-catalog-hide) saves `NO`, removes the card, and deletes the chat with its history. Publishing again creates a new, empty chat.

[GET /v1/infra/servers/:id](/docs/infra/servers/get) and [GET /v1/infra/servers](/docs/infra/servers/list) return `showInBitrix24` with the value `NOT_CHOSEN`, `YES`, or `NO`. The access policy, granted permissions, and app links are preserved.

### NEW-1117-8: deals in CRM API v3 mode with the X-Crm-Api header

CRM routes accept the `X-Crm-Api: v2|v3` header. In `v3` mode the deal routes [GET /v1/deals](/docs/entities/deals/list), [GET /v1/deals/:id](/docs/entities/deals/get), [POST /v1/deals](/docs/entities/deals/create), [PATCH /v1/deals/:id](/docs/entities/deals/update), [DELETE /v1/deals/:id](/docs/entities/deals/delete), [POST /v1/deals/search](/docs/entities/deals/search) and [POST /v1/deals/aggregate](/docs/entities/deals/aggregate) work through the new version of the Bitrix24 REST API with the same paths and field names. Whatever `v3` mode cannot do is refused with `422 CRM_V3_UNSUPPORTED` and `error.details.retryWith`, a hint to repeat the same request with the `X-Crm-Api: v2` header. The same refusal comes back when the Bitrix24 account does not have the new version of the Bitrix24 REST API. The response names the mode in the `X-Crm-Api` header when the request sent that header or the key works in `v3` mode. [GET /v1/me](/docs/keys-auth/me) returns the `crmApi` block with the key mode. Existing keys stay in `v2` mode, and their requests without the header work as before. The mode in which a Bitrix24 account issues new keys is shown by the `crmApi.newKeyMode` field. Details: [CRM API modes](/docs/crm-api-modes).

### NEW-1117-9: Agent maintenance key delivery by code

When maintenance key delivery is enabled on the platform, `GET /v1/me` with a code describes `kind: "agent_maintenance"`, and `POST /v1/key-claim` returns one key in `VIBE_MANAGEMENT_KEY`: JSON with `api_key`, `env_name`, `key_fingerprint`, `key` or private dotenv with a final `# end`. Availability depends on platform enablement.

The recipe saves the key only to the servicing agent's local `.env`. Verification uses local SHA-256 prefix16 and one request to `GET /v1/infra/servers/:serverId/logs?service=hermes&lines=1` for Hermes or `GET /v1/infra/servers/:serverId/logs?service=docker&lines=1` for OpenClaw, requiring HTTP `2xx`, `success: true` and nonempty `data.logs`. `/v1/me` with the maintenance key itself still returns `403 AGENT_MAINTENANCE_KEY_OUT_OF_SCOPE`; verification failure leaves the saved key intact and does not require revocation. Application, server and MCP environments stay unchanged. See [exchange and recovery](/docs/keys-auth/me), [MCP](/docs/mcp), [infrastructure](/docs/infra).

### BC-1117-10: nested deal, lead, contact and company routes in CRM API v3 mode

> Old format supported until: not provided

**Before**

In `v3` mode, moving a record to another stage and the links between records (deal, lead and company contacts, a contact's companies) ran through the previous version of the Bitrix24 REST API, as in `v2` mode. The deal contact list returned `sort` and `roleId`, and moving a deal to another stage returned the deal with all fields of the previous version.

**After**

In `v3` mode, [POST /v1/deals/:id/move](/docs/entities/deals/move) and the deal contacts routes ([GET /v1/deals/:id/contacts](/docs/entities/deals/contacts-list), [POST /v1/deals/:id/contacts](/docs/entities/deals/contacts-create), [PUT /v1/deals/:id/contacts](/docs/entities/deals/contacts-set), [DELETE /v1/deals/:id/contacts/:relatedId](/docs/entities/deals/contacts-delete-delete) and [DELETE /v1/deals/:id/contacts](/docs/entities/deals/contacts-delete)) run through the new version of the Bitrix24 REST API. Moving a deal returns the deal without the fields that have no counterpart in the new version. In the new version, the link between a deal and a contact has no sort order and no role: the contact list comes without `sort` and `roleId`, the primary contact is first, and `sort` or `roleId` in the request body is refused with `422 CRM_V3_UNSUPPORTED`. Linking a contact with `isPrimary: true` takes two Bitrix24 calls and is not atomic: if the second call fails, the contact stays linked but not primary.

Moving a lead to another stage [POST /v1/leads/:id/move](/docs/entities/leads/move), lead contacts [GET /v1/leads/:id/contacts](/docs/entities/leads/contacts-list), company contacts [GET /v1/companies/:id/contacts](/docs/entities/companies/contacts-list) and a contact's companies [GET /v1/contacts/:id/companies](/docs/entities/contacts/companies-list), together with their write and unlink operations, respond in `v3` mode with `422 CRM_V3_UNSUPPORTED` and `error.details.retryWith`, a hint to repeat the request with the `X-Crm-Api: v2` header. In `v2` mode all these routes work as before.

**What integrators should do**

If you need `sort` and `roleId` of deal contacts, the deal with all fields of the previous version in the move response, or the nested lead, contact and company routes, a key in `v3` mode sends the `X-Crm-Api: v2` header on these routes. Details are in [CRM API modes](/docs/crm-api-modes) and in the "v3 mode" section of the [Deals](/docs/entities/deals) page.
