Your workspace, as a live process graph.

Pulse reads the everyday stream and pulls out the decisions, commitments, features, past failures, and customers, each with a note on how sure it is.

• ▸Structured, not blobs. Every item Pulse saves has a type, a reason, and an owner, so you can actually look it up later.
• ▸Only when it’s sure. If Pulse isn’t confident enough, the item stays admin-only and is never shown as fact.
• ▸You’re always in charge. Anyone can edit, archive, or split anything Pulse captured, from any screen.

Pulse links what it found: which decision led to which commitment, who owns which feature, and the sourcebehind each one. Your company’s map is walled off from every other company’s.

• ▸Typed links, not guesses. A dozen relationship kinds (decided_by, owned_by, implements, evidence_for, references, shipped_for, expert_in, threatens, and more) tie each item to its source and its consequences.
• ▸Walled off per company.Your map can never touch another company’s; the isolation is structural, not a setting. BYOK on Enterprise.
• ▸Always traceable. Every link points back to the exact source event behind it, so you can check why two things are connected.

Pulse watches continuously for stuck PRs, stalled decisions, churn risk, and cross-team gaps. It only speaks up when something needs you, never a firehose.

• ▸No nagging. Once Pulse flags something, it won’t flag the same thing again for a set window.
• ▸Goes to the right person. Pulse notifies whoever owns the work, not a whole channel.
• ▸Admins see it all. Founders and admins get a calm daily digest of everything Pulse flagged across the company.

When something needs doing, Pulse drafts it: the Slack DM, the Linear ticket, the calendar invite, all waiting in your approval inbox. One click sends, and you have five minutes to undo. Drafts come from three places: simple rules, multi-step playbooks, and custom agents you build with a trigger and a prompt, no code.

• ▸Playbooks that branch and loop. A playbook can take different paths (enterprise customers escalate) and repeat a step for each person (DM everyone on call), not just run top to bottom.
• ▸Custom agents in two minutes. Pick a trigger, write a prompt, choose what it proposes, and decide whether a human approves. Start from one of 20 templates across sales, marketing, support, HR, legal, and ops, or from scratch.
• ▸Each agent reads only what you can. An agent reads only what its creator can see, set by an explicit grant (a team, channel, source, or doc); no grant, no reads. Scope filters narrow the subject it searches, but the grant is the permission boundary.

Tap the voice button on any page and just talk. Pulse answers out loud with citations, reads you a briefing, or drafts a Slack DM or Linear ticket you approve by voice. Every spoken action runs through the same ACL gate and the same five-minute undo as everything else.

• ▸Same engine, spoken. Voice uses the same retrieval and citations as text; sources are read out and shown in the transcript, never invented.
• ▸Approve out loud. Any draft Pulse makes by voice waits for a spoken yes, and the five-minute undo still applies to every send.
• ▸Five briefing types. Morning, evening, pre-meeting, topic, and person briefings, hands-free on the commute or between meetings.