Agent runtimeagenticopsresearchgtmcodingdocumentedunattended capable

Operational brief

Hyperagent

Hosted agents for attended research and building, with optional schedules, Live Mode and event-driven work under workspace and knowledge controls.

No product visual on file

Choose it when

You want to shape work in an attended thread, then add a bounded trigger and deliberately widen its tool, account and knowledge access.

Skip it when

You need to host the execution environment yourself or require independently verified universal approval and retention controls before sharing sensitive systems.

01 / Classification

Autonomy evidence

This is our classification, not a vendor badge. The record says what the product normally does. The gaps stay visible.

How we classify it

Tool use + multi-step plans, still human-gated per session.

Runs unattended

Yes — that capability is documented in this brief.

Trigger

Docs start with attended threads, then optional schedules, Live Mode, channels, webhooks and agent handoffs. Named-agent invocations add unattended capability; schedules default to read-only and trust is widened per doorway.

State

Each run has a thread; invocation configuration chooses fresh or continuing threads. Knowledge read scope and save destination are independent. Workspace agents use workspace/agent knowledge rather than the runner's personal memories; external triggers cannot use thread search.

Tools

Browser, code/files, research and publishing tools do real work. Code execution is always available. Supported scheduled/Live Mode writes to connected apps can act, stay read-only or ask; that control isn't a blanket browser/shell approval policy.

Human control

Each invocation has an owner whose departure pauses it. Users must review questions, approvals and failing schedules in Inbox and investigate failures, not just clear notifications. Workspace Owners oversee the fleet; Editors maintain shared resources and Members run them.

02 / Operations

Control reality

Useful agents eventually touch something real. This is the operational picture we have, not a promise hidden in a sales page.

Where it runs

Hosted computing environments and built-in browsers/code tools; not a local runtime or a self-hosting assurance.

Data boundary

Privacy policy permits Content processing by third-party AI providers. Deleted content may remain archived or backed up; permanent deletion has a request path. Personal threads/memories aren't automatically shared, but Owners have oversight of members' workspace threads.

Identity & permissions

Member run access, Editor maintenance and Owner administration are distinct. A shared agent may use its owner's account or each user's own connected credentials; sharing alone doesn't choose execution identity. Invocation runs use the configuring owner's context and usage.

What it can write

Hosted webpages and experimental applications · Documents, decks, code and files · Supported connected-app actions under invocation-specific write controls

Surfaces

Web command center · Slack · Telegram · Webhooks · Scheduled runs · Vendor-provided compute environments

Export / lock-in

Outputs can be published as shareable webpages and slides. Complete workspace/agent export portability isn't established here; the privacy policy offers data requests and a permanent-deletion contact, not a proven round-trip runtime export.

03 / Receipts

Evidence ledger

Here is what this brief is built on. “Documented” means we found a source. “Tested” means this desk actually ran it. Those are different things.

Evidence status

documented

Source set

11 primary sources

Evidence review

2026-10-15

Sources on file

Corrections

2026-10-15 · Primary facet now reflects attended-first work with optional invocations. Added scoped connected-app write controls, owner/credential and workspace oversight boundaries, Inbox failure responsibility and privacy retention limits. Vendor learning case study remains attributed, not desk evidence; not runtime-tested. Release metadata retained rather than treating a docs review as a release.

Desk notes

What we found

Useful context, not a substitute for the source record above. This is not a desk test unless the evidence status says tested.

Pick it for hosted multi-tool work you can refine before automating; skip it for assumed universal approval or local-only data control.

Normal start

Attended thread before optional triggers

Writes

Supported app writes: act, read-only or ask

Governance

Members run; Editors maintain; Owners oversee

Failure owner

Invocation owner and user-reviewed Inbox

What the vendor says

Docs separate invocation trust, knowledge scope and credentials. The vendor's sales-pipeline learning story isn't our experiment; not runtime-tested.

Sources fetched 2026-10-15

04 / Change log

Related coverage

The brief tells you what the product is. Payloads track the changes that may alter the decision.

No linked change coverage on file yet.