update_trigger API rejects any model change on scheduled Routines (model_update_disabled)

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Claude Code Remote's scheduled trigger ("Routine") API exposes a model field on update_trigger, but calling it always fails with model_update_disabled — regardless of which model is requested, including re-setting the trigger's current model. This makes it impossible to run a scheduled/recurring trigger on a specific model once created; every fresh-session trigger silently falls back to the environment default.

What Should Happen?

Either update_trigger's model parameter works as documented, or it's removed from the tool schema so it doesn't appear usable when it isn't.

Error Messages/Logs

failed to update trigger: updating a routine's model is not available (model_update_disabled)

Steps to Reproduce

Create a scheduled trigger via create_trigger (no model param — it isn't accepted at creation either).
Call update_trigger with trigger_id set and model: "claude-sonnet-5".
Observe: failed to update trigger: updating a routine's model is not available (model_update_disabled)
Repeat step 2 targeting model: "claude-fable-5" (the trigger's actual current/default model) — same error. This confirms it's not a specific-model restriction, it's that the model field can't be changed at all via this endpoint.

Claude Model

None

Is this a regression?

No, this never worked

Last Working Version

N/A — Claude Code Remote (scheduled Routines), not local CLI

Claude Code Version

2.1.220 (Claude Code)

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

Impact: Environment default model for scheduled triggers appears to be an older model (claude-fable-5) rather than the account's preferred model (claude-sonnet-5), with no way to override it per-trigger. For production automations running hourly/daily on real business data, this means every scheduled run gets weaker reasoning than an equivalent interactive session, with no workaround short of binding the trigger to a persistent session (a different, riskier architecture for long-running production jobs).

Related existing reports (same trigger/Routine system, different symptoms):

• #51151 — RemoteTrigger API "Unable to resolve organization UUID"
• - #44811 — /schedule skill fails to connect to remote claude.ai account
• - - #30649 — Feature request for scheduled/cron task support