Help

Run your first matter task

Type what you need on Home — a draft, a memo, a research question — and send it. The matter team works in front of you: the roster shows who is working, and the run settles into a document or an answer you can export.

Chat mode gives you the Senior Associate's first pass; Draft mode adds the full review bench (issue-spotting, risk, adversarial review, verification, citation re-check); Research mode is source-forward — one Senior Associate pass with live authority gathering and citation-focused output, and no review bench.

Each mode runs on the engine your workspace is configured with — a fixed setting for this deployment, not chosen per run.

Inbox — attention, review, and scheduled work

Inbox is one surface with three views. All shows what needs your attention now and your recent activity; Review is the full queue of everything awaiting your judgment; Scheduled holds your Keep-Watch monitors and the sweep history. Nothing is ever sent from this surface.

The view you are reading lives in the address bar, so a reload, a bookmark, or a link you paste to yourself reopens the same view. Counts come from your real records — a badge only ever shows work that genuinely awaits you.

Matters & the Command Center

A matter is the durable container for a piece of work. Create one from the Matters surface (or the rail's New matter), and open it to see its brief, runs, documents, timeline, and audit trail in one place.

Rename or archive a matter any time — archiving parks it (reversible), it never deletes work.

Status labels & what verification means

Every document Talden files is marked DRAFT and "Not verified". DRAFT means it needs your review before use — Talden never produces finished work product on its own. "Not verified" means no citation or authority check has run over that document, so Talden makes no claim about it.

Talden never marks its own work product verified. The Statement-of-Law reviewer flags what it cannot confirm; it never asserts an authority is good. A label that stays "Not verified" is the guardrail holding, not a feature that failed.

When the drafting provider errors mid-run, Talden files a labeled skeleton instead of text it can't stand behind: those documents read "Fallback draft — provider error" and the matter reads "Work product degraded — re-run needed". Re-run the matter once the provider is available.

Library & exports

Every settled run records to your Library when the durable store is on — browse, preview, and search your whole practice, then export any recorded turn to Word or PDF, with cited authorities appended when present.

When Microsoft 365 is connected in Settings › Integrations, Library rows can also save to OneDrive or open as an Outlook draft — a draft you review and send yourself; nothing is ever sent for you.

Playbooks

Playbooks are what the matter team knows how to do. The Official tab is the built-in library the Orchestrator Associate assigns per matter — the playbooks layer additively onto each specialist's role, and there is nothing to toggle.

The Personal tab holds playbooks your own runs proposed. Nothing is auto-promoted: a candidate advances only when you decide it, and approving records your disposition — the runtime library itself changes only by an explicit code change.

Orchestrating Agent

The Orchestrating Agent is your account-level supervising agent — one page for what is running across your matters, what awaits your approval, and the verifiable record: matter triage, the agent roster, standing orders, the memory agents may draw on, connectors, proposed playbooks, and the audit timeline.

It proposes; you approve. Nothing executes while an approval is pending, a standing order you issue is carried into later runs on the matter you choose, and every decision is recorded to the append-only audit log.

Review — approvals in your Inbox

Review is a view of your Inbox: one queue for everything awaiting your judgment — approval gates, proposed memory, proposed playbooks, flagged verifications, and matters needing review.

Nothing executes while a gate is pending — approving is what lets an action proceed, and both decisions are recorded in the audit trail.

Scheduled — Keep-Watch in your Inbox

Keep-Watch monitors re-check a research question on a schedule and raise a signal only when the sources genuinely change. Add or remove watched queries in Inbox › Scheduled.

The view tells you honestly whether the daily sweep is actually scheduled in this deployment — a saved item never implies a sweep that can't run.

Data Controls & retention

Data Controls lives in Settings — open the Settings menu at the foot of the rail, or the Settings button in the Home header on smaller screens. When the durable store is configured it operates the real retention posture: purge your everyday history, purge a single matter (typed confirmation, one audit event), and download the hash-chained audit record as a CSV with its integrity verdict.

Attachments you upload for a run are used for that run and not kept; recorded work persists only when the durable store is configured, and the surfaces always disclose which is true.