Services
Your Stone Garden workspace.
Loading…
Use your company Microsoft account.
Stone GardenPortalYour Stone Garden workspace.
Loading…
Use your company Microsoft account.
Create, publish and share interactive sites.
Your systems, their care, and the agent behind them.
Activity, run history, and maintenance results.
Waiting for worker reporting…
Waiting for the office worker.
Whole PC totals, including other applications.
| Job | Submitted by | Priority | Status |
|---|
Routines, review evidence and work to follow up.
Result feeds retain observation times and turn stale explicitly. Agent reviews and supporting automated evidence are reported separately.
Imported evidence only. Missing receipts do not establish missed reviews.
Recorded findings and next actions. Dates refer to evidence, not continuous health.
Project, instructions and approved access scope.
Source hashes identify the reviewed Agents checkout. The dedicated Agents project is the maintained source. A published file does not prove a session loaded it.
Manage access to your services.
Removing access keeps saved canvases.
| Employee | Canvas | Action |
|---|
Service access and site sharing are separate. Removing access preserves authored content.
Quarantine withdraws content. Removing quarantine leaves the site unpublished.
20 sites per author · 20 MiB and 1000 files per version · 100 MiB per site · 500 MiB per author.
Set job priority and resource limits.
Compute access is separate. Critical priority is unavailable.
| Employee | Compute | Limits | Action |
|---|
AI calls use the shared API account and incur charges. Removing access keeps saved request results private.
| Employee | OpenRouter | Action |
|---|
Read-only benchmarks and cost estimates. Uses a shared free API quota.
| Employee | Artificial Analysis | Action |
|---|
Ask a model about text, images, video or audio.
Compare model capabilities, speed and cost.
| Model | Intelligence | USD / 1M tokens Input / output | AA benchmark USD / task | Output tokens / sec |
|---|
Token-volume estimate, separate from AA benchmark cost per task. No model is run.
Indices compare benchmark performance; they are not universal percentages. Missing values are unmeasured, not zero. Cost per task describes AA’s evaluation workload, not your project or a guaranteed successful result.
Estimates assume uncached input and all billed output, including reasoning. Tools, retries, provider discounts and tax are excluded.
Source: Artificial Analysis · Free API · Shared cached snapshot.