Compute
Changelog

Changelog.

What shipped.

  1. CLI 0.1.18 · Reliability

    Inspect startup failures and uncertain cleanup

    Startup evidence stays with the run, and uncertain provider cleanup remains visible.

    • Use compute logs RUN_ID --format jsonl, or compute_get_logs with format=jsonl, to inspect startup observations even when application logs are empty.
    • Provider status, reported container uptime, agent readiness and workload start are recorded separately. Missing or zero uptime is not proof that your code ran.
    • If an allocation response is lost, cleanup stays unconfirmed until the provider identity is reconciled. The release adds guarded monitoring delivery; live routing and exact-target GPU qualification are tracked separately.
    • Reinstall from the quickstart for CLI 0.1.18. Existing clients remain supported.
    Reinstall the CLI
  2. CLI 0.1.12

    Report a problem from the CLI or an agent

    compute report files a bug report with a durable acknowledgement id, and answers at once when the cause is already known. Reinstall from the quickstart to get CLI 0.1.12.

    • compute report "<summary>" --run run_… stores the report before returning rpt_…, and attaches the run's status, provider, and error so you do not have to.
    • Outdated CLI, low balance, a known provider condition, or a duplicate of your own open report get an immediate reply; everything else is answered by an operator, visible with compute report status rpt_….
    • Agents get the same through the compute_report_issue and compute_get_report tools in compute mcp and mcp.compute.cx.
    Reinstall the CLIRead the docs
  3. Operations

    Safer website releases and status checks

    Website releases now verify production authentication before going live, and status checks no longer depend on homepage wording.

    • The outage happened because one release bundled Clerk's test publishable key while production used a live secret key, so Clerk rejected the mismatch.
    • Compute's website is built in its production environment and rejects a release if Clerk points to the wrong application.
    • The Website component now follows HTTP availability. Authentication keeps separate checks for the production Clerk origin and sign-in page.
    Check current status
  4. CLI 0.1.11

    One MCP tool catalog

    Local and hosted MCP now expose the same tools. Reinstall from the quickstart to get CLI 0.1.11.

    • compute_credits_add now takes amount_cents. Update clients that pass amount_dollars.
    • compute tinker connect and tinker start --print-exports were removed. Use compute tinker run, exec, or shell to keep the SDK key in the child process.
    • The smaller CLI no longer includes unused provider code and only connects to api.compute.cx.
    Reinstall the CLIRead the docs
  5. Operations

    Public status page

    Check Compute's health and 90 days of incident history without signing in.

    • The website, API and CLI, and MCP endpoints are checked every 30 seconds.
    • Authentication, GPU jobs, deployments, and billing have separate status components.
    View the status page
  6. CLI 0.1.10

    Let Compute choose the GPU

    The CLI can choose a GPU by speed or price. You can still name a specific card.

    • Use auto for the default choice, fast or fastest for speed, and cheap or cheapest for price.
    • Dry runs show the paid step that comes next without launching a machine or spending credit.
    Open the quickstart
  7. CLI 0.1.8

    Use Compute from MCP clients

    The compute mcp command exposes account, GPU, dry-run, run, log, result, and cancellation tools through MCP.

    • Set confirm_spend=true to start paid work.
    • Secret values are never returned. COMPUTE_NO_TELEMETRY disables MCP and CLI telemetry.
    Read the docs