Sign in with IAM.Identity and connect Speech resources

Staging candidate, not a production release. This flow was tested at https://app.stage.iamspeech.ru with a separate synthetic QA profile. An operator enables sign-in and IAM.Core delegation; a published tool does not mean your account already has a configured connection.

Sign in and open the workspace

1. Open the IAM.Speech sign-in page. Select RU or EN.
2. Click Sign in with IAM.Identity. Enter your Identity password only on id-staging.iamrouter.ru for this environment.
3. Complete normal IAM.Identity sign-in. If new terms or consent appear, review them and decide yourself whether to continue.
4. After returning to Speech, Call recordings opens. The synthetic test created a separate workspace with the operator role and a zero-minute balance. An empty list means this workspace has no recordings.

Initial workspace creation requires an exact operator grant. Matching an existing Speech account email does not merge accounts or grant another company’s access. Identity groups and roles do not make a user a Speech administrator. The company is the workspace; this change does not add separate project CRUD.

If Sign-in could not be confirmed. Start a new sign-in; this exchange will not be repeated. appears, click Return to sign-in and start a fresh normal sign-in. Do not resend an old callback link or code. For an initial access refusal, ask the operator to check the grant for your exact profile. The short session requires normal sign-in again after expiry; automatic renewal and immediate revocation of every local session upon Identity logout are not claimed.

Enlarge screenshot

1

1. Sign in with IAM.Identity

Enlarge screenshot

1 2 3

1. Call recordings
2. Zero minutes and zero completed
3. Owned synthetic queued record

Read and add resources through IAM.Core

The candidate IAM.Agent integration uses its configured IAM.Core MCP server. An operator binds one exact Core tenant/principal to the current Speech user and company. Tool arguments cannot select another company.

Available reads are speech.dialogues_list, speech.dialogue_get, speech.dialogue_segments, speech.dialogue_metrics and speech.dialogue_activity. A separate speech:write grant enables speech.dialogue_create and speech.dialogue_upload. Create and upload are separate actions with an explicit reason and approval; listing records alone never authorizes a write.

For Core uploads, supply the exact UUID of your dialogue, a safe filename, MIME type and at most 8 MiB of audio. A short synthetic IAM.Voice-generated WAV was actually tested; this run does not qualify all other advertised codecs. A saved file or accepted start is not completed analysis. Check the stored status and actual segments, metrics and activity. Re-read the record before reconciling an unknown upload outcome; never automatically upload it again.

The synthetic test returned real transcript segments, metrics and activity. The record remained processing; completed processing and minute debit with a zero balance were not confirmed. In the first phase, the short SSO session limited static operator-held Core delegation.

The newer staging version provides a separate limited IAM.Agent connection form with recipient/permission review and consent lasting at most 15 minutes. Request creation and resource refusal before consent were checked live; consent itself, access refresh and the new connection’s resource operations still await live qualification. Ordinary sign-in, the former static delegation and tool availability do not replace that consent or establish permanent access.