Provider accounts and Fast mode

List the model accounts connected to a Workspace, name them, and run a turn in Fast mode over the API.

A Workspace can connect several accounts of one provider, for example three ChatGPT subscriptions. The API lists them, names them and picks Fast mode, as the model picker does in the app.

​

What you will have

• Each connected account’s key, email and name.
• An account renamed, so people and agents can tell it apart.
• A Session turn running in Fast mode.

​

Prerequisites

An API key from Settings > API with org:read to list accounts, org:write to rename one, and run for Sessions. Keep your Workspace id from GET /v3/self.

export REASON_API_KEY=reason_...
export ORG=your-org-id

​

List the connected accounts

curl -s https://api.reasonmachines.com/v3/organizations/$ORG/provider-credentials \
  -H "Authorization: Bearer $REASON_API_KEY"

Each item names one account:

{
  "provider_id": "openai",
  "account_key": "chatgpt-59f7c8fc52f526d05fc10c762ac88ed3",
  "type": "oauth",
  "account_email": "[e-mail]",
  "account_name": "Your Name",
  "account_display_name": null,
  "auth_failed_at": null
}

account_email is the email the provider reports for a subscription sign-in. account_display_name is the name someone gave the account. Secrets are never returned.

​

Rename an account

curl -s -X PATCH \
  https://api.reasonmachines.com/v3/organizations/$ORG/provider-credentials/openai/accounts/chatgpt-59f7c8fc52f526d05fc10c762ac88ed3 \
  -H "Authorization: Bearer $REASON_API_KEY" \
  -H "content-type: application/json" \
  -d '{"display_name": "Work ChatGPT"}'

The name replaces the email in the model picker and in Settings > Models. Send null to clear it. Names are trimmed and cut to 80 characters. A member renames only accounts they connected; owners and admins rename any. Reference: Rename a provider account.

​

Run a turn in Fast mode

Add speed: "fast" when you create a Session, send it a message, or message a Reason:

curl -s -X POST https://api.reasonmachines.com/v3/organizations/$ORG/sessions \
  -H "Authorization: Bearer $REASON_API_KEY" \
  -H "content-type: application/json" \
  -d '{"prompt": "Fix the failing test", "model": "byok:openai-codex:gpt-6.1-sol", "speed": "fast", "idempotency_key": "fix-test-1"}'

Fast mode asks the provider for its priority tier, which answers sooner and uses the plan faster. Today that is a ChatGPT subscription’s models; on other models the turn runs at standard speed. To pick one account, add its model_source_id from List models (groups[].accounts[].source_id).

​

Rules that surprise people

• speed applies to the turn its message starts: a new Session, a finished Session’s next message, a queued message (delivery: "queue") or a Reason’s first message. A message steered into a running turn joins it at that turn’s speed.
• Each turn names its own speed. Omit speed and the turn runs at standard.
• A Session started in one of the app’s capability modes keeps standard speed unless the message also sends model or reasoning_effort.
• Any value other than standard or fast returns 400.