Pulse
Welcome back
Identity-only sign-in. No workspace-level permissions asked — we just need to know it's you.
Sign in with SlackWorkspace hasn't installed Pulse yet? Add to Slack →
Pulse is built around the moments that quietly hold a team together — the weekly briefing, the casual intros, the watercooler question, the noticing of someone who's been quiet. Each one is a proper feature with its own safety rails, opt-outs, and analytics.
Create a pairing program per channel or audience — Engineering mixer, all-hands, designers, whatever. Each program has its own schedule, group size (pairs or triads), no-repeat window, and audience scope.
When a round runs, Pulse scores every legal partner for each person by shared topic tags (55%), complementary activity levels (25%), and freshness of engagement (20%). Then it writes a "why this pair" brief and a conversation starter tailored to both profiles. Three days after a pair marks "we met", Pulse asks a one-tap "how'd it go?" — winning starters feed the next round's AI prompt.
Try pairings free →Each watercooler is bound to a Slack channel — General, Project, Guild, or Random — and the AI matches the tone. A project-thread gets energising questions about craft; a random channel gets playful; a guild channel gets thoughtful. The bot posts, then drops its own first reply 60 seconds later to break the ice.
AI proposes fresh questions from what the channel was actually talking about that week — every suggestion shows permalinks to the messages that inspired it, so admins approve in context. Humor guidelines are hard-coded: no sex, stereotypes, politics, mental health, health, family, finances, performance, or trending crises. When in doubt, kindness wins.
Try watercooler free →When someone replies in a watercooler thread, Pulse can write a short personalized follow-up — either in the same thread, or as a private DM if they'd engage more there. The AI reads their past replies for context, but paraphrases, never quotes.
Every user can type "show my profile" or "delete my profile" as a Slack DM. Consent is a hard gate, not a checkbox. Rate limits cap DMs at one per person per week.
Read the safety rails →Here's what I have:
"Short read: Often writes about travel and photography. Tends to answer longer than average; engages with prompts about craft."
Tags I picked up: travel, photography, film, cooking.
Reply 'delete my profile' anytime to wipe this.
Create a newsletter with a title, context, and objective. Pulse suggests questions (using past-run performance as signal) — you approve, edit, or write your own. Every question becomes its own DM thread, so a teammate answers one, then moves on.
Each greeting is written fresh for that person using their first name, their inferred communication style, the time-of-day, and the objective. Follow-ups reference what they just wrote, not a template. When the window closes, a consolidated digest publishes to the channel you chose.
See a newsletter →Apr 14–18, 2026 · 14 teammates replied
New onboarding flow went live Tuesday. Three client-facing dashboards rolled out across APAC. Timezone-aware scheduler is production-tested on a 150-person workspace.
Stripe migration waiting on legal review (early next week). Infra ticket on the queue worker still open — owner on leave, backup picks it up tomorrow.
Q3 planning doc. Two folks prototyping an HRIS sync. Interest in an in-person team retro — seems real.
One page answers the questions a People Care lead actually asks: Are people participating? Are pairs meeting? Which prompts landed? Who's been quiet for 60+ days?
Each metric is colored green/yellow/red against a healthy baseline — not a vanity number. The dormant-users widget lists people with zero signals in 60 days; a one-click "send a thoughtful nudge" action writes a DM that references their own known interests, not a template.
See a live dashboard →Privacy and consent aren't checkboxes at the bottom of an admin form. They're enforced in code — no AI can override them, no dashboard can bypass them.
Every teammate has four flags: pairing_opted_out, watercooler_opted_out, dm_follow_ups_opted_out, ai_profile_opted_out. Each is a gate in the service code — nothing sends when the flag is true, regardless of what the AI decided.
Type pause, resume, snooze 2 weeks, don't pair me with @person, show my profile, delete my profile, or integrations as a Slack DM to the bot. No admin dependency.
Every AI agent is explicitly forbidden from quoting past replies verbatim — paraphrase only. Your replies are never used to train external models.
Hard-barred in every agent's instructions: health, family, finances, religion, politics, performance reviews, sex/gender, stereotypes, trending crises.
Max 1 DM follow-up per user per 7 days. Max 1 thread follow-up per user per watercooler round. Post-meeting feedback fires once per pair, on the transition.
Every LLM call writes a row to ai_traces — agent name, model, tokens, cost, latency. One-row-per-call auditability, not aggregated.
AI_DEFAULT_PROVIDER to swap. Every call is audited.getSchedule the same way, for Outlook-native teams.14-day free trial. No credit card. Install from Slack, turn on the rituals you want.
Start free trialDonut is a solid product. Here's where Pulse is different — and where we're not.
Comparison reflects Pulse features that are live in the codebase today. Donut rows reflect their published documentation at the time of writing. We'll update this as either product changes.