Summary
As of October 2026, official documentation describes several ways to automate Claude cloud work, but dispatch acceptance does not prove completion or promotional-credit eligibility. Public issue reports show that trusted-device rejection can persist despite reported enrollment, while Reddit billing accounts remain anecdotes rather than authoritative promotion terms. [1][3][4][5][6][7][8]
Method and collection date
This report synthesizes public research collected on October 21, 2026: Reddit discussions, public GitHub issues and repositories, then official documentation. Exact-error searches did not locate a directly relevant Reddit report; that is a search limitation, not evidence that none exists. No account access, login, task dispatch, installation, scheduler change or billing measurement was performed. Retained URLs received scripted HTTP availability checks on November 2, 2026. These checks do not revalidate document contents, issue resolution or promotion terms.
Findings
Device enrollment and successful authentication are not equivalent
A public October 4 issue describes trusted-device rejection when forking a headless Linux Remote Control conversation from Desktop, even though the reporter says enrollment and device attestation are present. It was open when collected in October 2026, with no verified fix visible in the inspected material. This is a user report about a related operation, not a reproduced diagnosis of every --cloud failure. [1]
The Remote Control documentation inspected in October 2026 distinguishes device enrollment from recent sign-in, with recent authentication aging after 18 hours. It describes enrollment during /login, with individual or organization control depending on the plan. It also says setup-token credentials cannot establish Remote Control. These contracts concern Remote Control; applying them to cloud creation is a diagnostic hypothesis, not a proven fix. [2]
The defensible conclusion is narrow: a trust-related rejection need not mean enrollment is missing, and a model-request token is not interchangeable with full sign-in for the documented Remote Control flow. No particular CLI release or flag was established as a universal remedy. [1][2]
Execution location matters more than similar-looking flags
The following contracts were documented as of October 2026. They should not be assumed to exist in every older CLI version. [3][4]
| Interface | Documented behavior | What it does not establish |
|---|---|---|
--cloud | Starts cloud work. [3] | Successful trust verification or promotional-credit charging. |
-p with --cloud and an existing cloud target | Queues a follow-up and exits; documented for scripts and CI. [4] | A completed reply, completed artifact or exemption from authentication checks. |
Plain -p | Noninteractive CLI execution. [3] | Cloud execution merely because the command is unattended. |
--remote-control / --rc | Remote steering of execution that remains local. [2][3] | Cloud computation or promotional cloud-credit use. |
| Routines | Scheduled, API and GitHub triggers for cloud work. [5][6] | Promotion eligibility from trigger acceptance alone. |
The cloud follow-up guide requires subscription sign-in and applicable organization policy. A send acknowledgement means the message was accepted for dispatch, not that the task finished. The reviewed material did not establish a minimum version for every documented follow-up behavior. [4]
As of October 2026, routine API/token setup is documented as web-only. The experimental fire API uses a per-routine bearer token rather than a Console API key and returns a cloud task identifier and URL. The documented token is trigger-only, not a general read credential. The reviewed material did not establish a general transcript/status polling API usable with that token. [5][6]
Promotional-credit behavior is not established by ordinary billing docs
The routine documentation inspected in October 2026 says runs consume subscription usage. That supports a billing statement about routines, not an assurance that a particular promotional-credit pool will pay for them. [5]
A September Reddit discussion reports promotion wording that regular plan usage resumes after promotional credit. It also includes a complaint that the displayed balance remained stationary while work had already fallen back to plan limits. This is a secondhand user account, including reported support advice, not direct official confirmation of fallback or display behavior. [7]
Another September thread describes bounded compiler-fix work reportedly costing single-digit to low-double-digit dollar amounts, with follow-up needed for patch handoff and cross-platform CI. These are uncontrolled self-reports, not a transferable cost estimate or a verified spending rate. No usernames or direct quotations are reproduced. [8]
The October research did not locate authoritative promotion-specific terms sufficient to establish current eligibility, expiry, balance accuracy or reliable consumption rates. Those questions remain unverified; no private account balance or expiry is needed to explain the public evidence. [5][7][8]
Queue software does not resolve authentication or billing uncertainty
As of October 2026, a public queue project advertises schedules, completion/failure callbacks and a remote mode that invokes the native cloud-launch command. That can organize work, but its description does not establish a workaround for native authentication failure. [9]
The official Claude Code GitHub Action provides repository automation through GitHub Actions and API/provider authentication. The reviewed material does not establish it as a dispatcher into promotional-credit-eligible cloud work. Neither project was installed or tested for this research. [10]
Limitations
Issue reports and Reddit discussions are anecdotal, and official documentation may describe behavior newer than an installed CLI. Device recovery, unattended dispatch and billing eligibility are distinct questions; success in one does not prove the others. No recovery was reproduced, no cloud result was observed and no credit meter was measured. Findings reflect October 2026 evidence, not a claim about current account behavior. A later HTTP 200 response establishes URL reachability only.
References
- https://github.com/anthropics/claude-code/issues/99463
- https://code.claude.com/docs/en/remote-control#trusted-devices
- https://code.claude.com/docs/en/cli-reference
- https://code.claude.com/docs/en/claude-code-on-the-web#send-follow-ups-from-the-cli
- https://code.claude.com/docs/en/routines
- https://platform.claude.com/docs/en/api/claude-code/routines-fire
- https://www.reddit.com/r/ClaudeAI/comments/1woijhq/250_in_cloud_session_credits_a_reset_like_chatgpt/
- https://www.reddit.com/r/ClaudeAI/comments/1woj0m5/free_250_cloud_credit_on_5x_max_plan_thanks/
- https://github.com/MH4GF/tq
- https://github.com/anthropics/claude-code-action