Skip to main content
live across 25 connectors

Your company brain.built for the agent era

Never re-explain a decision.

Your team’s decisions and hard-won lessons scatter across ten tools and one person’s memory. Pulse keeps them in one place and drafts the next step when something needs one. It’s permission-aware, so it only ever shows people what they already have permission to see.

  • SAML / SCIM
  • ACL-mirrored retrieval
  • BYOK
  • Audit log

If you couldn’t see it in Slack, you can’t see it in Pulse.

hidden tax / 01

Your team’s most important decisions live in ten tabs.

The cost is quiet, daily, and well-measured. Six stats, six honest citations, no spin.

10

apps the average knowledge worker juggles every day

Asana · Anatomy of Work
1,200×

times a day they toggle between apps and websites

Harvard Business Review · 2022
9%

of annual work time lost to reorienting after each switch

Harvard Business Review · 2022

and it compounds
8mo

for a new hire to reach full productivity on a team

Allied Workforce Mobility Survey
and it compounds
23min

to fully refocus after every interruption (they happen every few minutes)

UC Irvine · Gloria Mark
and it compounds
60%

of the workday spent on work about work, searching, switching, tracking down decisions

Asana · State of Work Innovation 2024
what changes when you have pulse

And here’s what changes when the team has Pulse.

A new hire joins

Without Pulse

8 months to full productivity

With Pulse

A 30/60/90 plan, plus instant answers from everything the team already knows

Outcome

Up to speed in 30 days, not 8 months

A customer call in 20 minutes

Without Pulse

Skimming five tools to rebuild the context

With Pulse

A one-page brief: account history, open promises, recent sentiment, the ask

Outcome

Prepared for meetings in minutes

“Did we already decide this?”

Without Pulse

Hours re-litigating in a meeting

With Pulse

Pulse answers in seconds, with the source for every line

Outcome

5 seconds, with proof

Status updates and follow-up tasks pile up

Without Pulse

An hour of copy-pasting between tools

With Pulse

Pulse drafts the updates, nudges, and tasks from what actually happened; you approve, it sends

Outcome

Updates and tasks, automated

A question hits on the way to standup

Without Pulse

It waits until you're back at a desk

With Pulse

Say it out loud; Pulse answers from the same cited team memory

Outcome

Ask anything, by voice

Your AI tools start every session from zero

Without Pulse

Re-explaining the company to each assistant

With Pulse

Claude, Cursor, and Codex plug into the same workspace memory over MCP

Outcome

Context that follows you everywhere

AI is making teams smaller. That makes memory more important, not less.

75% of knowledge workers now use generative AI at work, nearly doubling in a single year. The same teams that needed 30 people ship at 10. The context each person carries has expanded, not shrunk.

Microsoft & LinkedIn · 2024 Work Trend Index
team ai / 02

Individual AI makes one person faster. Team AI makes the whole company smarter.

Everyone is handing teams individual AI: a chatbot each person has to re-explain the company to. Pulse is team AI, the memory layer underneath every tool your team already uses. It’s the company brain everyone works from, instead of a hundred private ones.

Individual AI · everyone else
  • A separate chatbot for each person.
  • Every one starts cold. You re-explain the company each time.
  • What one person learns, the rest of the team never gets.
Team AI · Pulse
  • One shared memory every person and tool draws from.
  • Ask once, and the whole company has the answer.
  • It gets sharper every time someone uses it.
why pulse / 03

Three things no other tool does.

Most company search bolts a chat box onto a document index, then adds permissions later. Pulse is built the other way round: it understands your work first, and respects who can see what from the start.

01 / process graph

It's a graph, not a search box.

Pulse keeps track of what you decided, who's on the hook for what, and what broke last time, and how they connect. You don't search for "what did we decide about pricing." You go straight to it.

02 / proactive

Surfaces work before you ask.

Stuck items. Stalled decisions. Customers whose champion just left. Cross-team meeting gaps. Pulse watches for these continuously and only speaks up when one shows up. Never a notification firehose.

STUCKPR awaiting review · 18h
DECISIONstalled 60d · no execution signal
CHAMPIONleft Pulse HQ · 90d to renewal
03 / agent actions

Drafts the work, never sends it without you.

Linear tickets, Slack DMs, kickoff docs, all proposed in your inbox first. Nothing leaves Pulse until you approve it, and you have five minutes to undo.

proposed · linear.create_issue
Onboarding cost analysis · refresh
team: PLATFORM · priority: P2
surfaces / 04

Six surfaces, shaped to the role.

Founders, eng managers, PMs, ICs, CS leads. Each sees a different first screen, a different draft queue, a different briefing. Knowledge captured once shows up wherever the role needs it next: Home, Ask, Search, Map, People, plus the Voice agent on every page when your hands are busy. Planning, customers, and the manager second-brain live as tabs on Map and People.

trust surface / 05

See what the AI knows, and what it’s only guessing.

Cited passages get a lime underline. Inferred passages get a dotted iris underline. Hover any inferred sentence to see the AI's reasoning. Toggle off if you want plain text.

you › why did we go with three pricing tiers instead of two?

The team chose three tiers: Starter, Standard, and Pro1 after a sustained back-and-forth in #pricing-strategy across late October.2 The third tier was added primarily to anchor enterprise deals and create headroom in negotiation though this rationale was never written down explicitly.

Apoorv called the decision on Oct 283 with explicit reference to the Q2 flat-pricing failure case4, where mid-market customers consistently downgraded to free. A two-tier structure would likely have re-introduced that downgrade pressure, but no one made that argument in the thread.

✓ confident · 84%4 cited2 inferredfreshness: 30d

How to read it

Cited grounded in source
Comes directly from a Slack message, doc or PR. Click the footnote to open the source. This is what we observed.
Inferred AI-derived
A claim the model is making by joining the dots. Hover for reasoning. Treat with the same skepticism you'd give a smart but new teammate.
Honest confidence calibrated weekly
Confidence is re-tuned against your team's own thumbs-up/down feedback, per team and per topic. Right by your team's own measure, not a benchmark.
Freshness older facts fade
Pricing goes stale in 30 days. Mission lasts 5 years. Pulse leans on recent sources and flags a claim when it's getting old.
what it does for you / 06
agent actions

Drafts the work. You're still the one who clicks send.

Every action (Slack DMs, Linear tickets, calendar invites, Notion drafts) lands in your actions inbox first. Approve, edit, skip. Drafts expire in 24h.

  • Send limits. Only approved domains, channels, and teams. Blocked actions never leave the building.
  • Five-minute undo. Every send is logged and shows an undo bar that retracts the message, deletes the comment, or archives the page.
  • Quota meter."23 of 50 verified actions used this month"; the price line, made real.
  • Test before you turn it on. See how often a rule would have fired against your real data before it goes live.
proposedslack.dm
expires in 23h 11m
to@daksh.engaboutPR #4421 awaiting review · 19hpolicystuck-pr-pinger · auto
Pulse APP
Hey Daksh, your PR #4421 (auth-refresh-token) has been waiting on review for 19 hours.
I checked who's free: @reema.eng is online and reviewed similar code last week. Want me to nudge her?
✓ allowedrecipient domain @northwind.io matchespolicy id: ap_4f2
quota · 23 / 50 used
sent · slack DM deliveredundo in 04:23
skills

Turn how your team works into playbooks your AI tools can follow.

When the tenth person asks how refunds work, that’s not a people problem, it’s a missing playbook. Pulse spots the pattern and compiles one from the process graph, with the decisions and past failures that shaped it built in, as a portable SKILL.md. Load it into Claude, Cursor, or any other MCP-aware client.

  • Knows when a playbook is stale

    A daily cron embeds new entities and scores divergence over 14 days. Skills past 0.6 flip to DRIFTED + write a notification. The export warns clients the procedure has aged.

  • Earns trust before it runs on its own

    Pulse tracks how well each playbook does in practice. A step can require a strong track record before it runs automatically instead of asking you first.

  • Permission-aware compilation

    Same skill, different SKILL.md per role. Pulse infers per-step required teams from the source entities and redacts steps the viewer can’t see. Every ACL the source already has, preserved.

● ACTIVErollout_pricing_change/SKILL.md
92% success rate
---
name: rollout_pricing_change
description: Roll out a new pricing tier across docs,
  billing, and customer comms.
version: 1.0.0
track_record:
  runs: 12
  success_rate: 0.92
---

## 1. Confirm the decision is in Decision Memory
Pull the entry. If the rationale isn't crisp, ask the
decision-maker for one sentence; every artifact cites it.

## 2. Update pricing-doc-v3 in Notion
Replace the tier table; preserve the prior version under
`archive/`. Tag #pricing-strategy in the change log.
built from 1 decision, 2 failure cases, 1 checkpoint
outputs

Turn what Pulse knows into something you can share.

Say “Pulse, make me a board update on the auth migration”and generation runs in the background. You get a toast and a home-rail update when it’s ready. Or click Save as on any Ask answer. Or pick a template from /app/present. Source-grounded from the same process graph that powers Ask.

  • Three output types

    Presentations (decks for the board, weekly digests, customer escalations). Dashboards (live project, customer, person, topic, team views with refresh policies). Documents (handoff docs, meeting prep, decision records).

  • Backed by sources, not vibes

    Every section is built from your decisions, commitments, features, past failures, and meetings, with a source on each claim. No web search, no making things up. If Pulse can’t back a claim, it flags the section for review instead of inventing one.

  • Edit inline, regenerate per section

    Click any section in edit mode, change the text, save. Regenerate one section with an optional hint. Revert to the originally-generated body when you change your mind. Full edit log retained.

● READYstripe-migration-status.pptx · 8 slides
quality 86 / 100
COVER
Stripe migration · weekly status
Apr 14 to May 14, 2026
WHERE WE ARE
84% of traffic now routes through Stripe Billing
Recurly cutover scheduled 2026-05-22 · Dana owns the runbook
OPEN COMMITMENTS · 3
Cutover runbook · Dana Park · 2026-05-20
Finance reconciliation script · Adekunle · 2026-05-18
cited from 3 decisions, 3 commitments, 1 failure case
voice + text

Say it once. Get the deck.

Voice in the morning while you walk. Text from your laptop at midnight. Same result: a board-ready document, backed by sources, ready to share. Usually inside two minutes.

VOICE · command in
Pulse, make the Q3 board update on the auth migration
~90 sec later
5-slide deck · cited

Auth migration · Q3 board update

  • Where we are, 84% of traffic on JWT
  • What we shipped, RFC-042, passkeys, SCIM
  • Open commitments, 3 · with owners
  • Risks + mitigations
  • Cited from 14 sources
14 sources citededitable · re-runnable
TEXT · command in
Live dashboard for at-risk customers this quarter
~90 sec later
live dashboard · refreshes hourly

Customers · at-risk lens

  • $418k ARR at risk across 7 accounts
  • avg churn-risk score · 68 / 100
  • 3 champions changed in the last 14 days
  • Quillwork · Lumen · Halftone · Patchwork
9 sources citededitable · re-runnable
VOICE · command in
Draft the auth-migration handoff doc for Therese's departure
~90 sec later
structured handoff doc · 5 sections

Auth migration · handoff

  • 01, Decisions made (RFC-042, RFC-058)
  • 02, Open commitments (3) + owners
  • 03, Architecture choices + tradeoffs
  • 04, Known risks + mitigations
  • 05, Recipients + read order
22 sources citededitable · re-runnable
proactive

Pulse notices. Before your stand-up does.

Pulse watches your work around the clock and speaks up only when something needs your attention. Never a firehose, and tuned to your role.

decision · stalled

“Auth migration v3” hasn’t moved in 60 days

Decided Oct 28. No PRs or Linear issues mention it since. Pulse spotted the gap, DMed @apoorv, and suggested a retro.

stalled-decision check · daily
cross-team gap

Platform ↔ Growth haven’t synced in 30 days

Two shared customers, one cross-team decision in flight. Pulse drafted an agenda: top 3 customers, top 3 decisions, one prompt each. Dismiss or schedule.

cross-team gap check · daily
customer · champion left

Pulse HQ lost their champion · 90d to renewal

Daniel K. (champion · 14 months) left Pulse HQ on Tuesday. Their renewal is in 90 days. AE notified, retention playbook drafted in your inbox.

champion watch + save play
stuck · stage

Interview loop stalled at “debrief” for 6 days

Anna’s panel finished Oct 24. The debrief usually happens within 3 days; it has been 6. Pulse pinged the loop owner and flagged it.

stuck-stage check · every 6h
common answer · ready to pin

“How does annual billing work?” asked 8× this month

Same answer landed every time. Pulse marked it as the go-to answer, pending your sign-off. Once approved, it pins above future answers with a freshness countdown.

repeated-question check · daily
security · sensitive

Credentials posted in #all-engineering

Pulse’s security scan caught an AWS access key in a public channel. It wrote up the finding and notified the owners. It won’t fix it automatically. Your call.

security scan · 02:00 UTC nightly
how it works / 07

From your tools to a trusted answer in five steps.

Every part of Pulse runs the same path, with one ACL gate and one audit trail behind every answer.

step by step · per event

Something happens in one of your tools. Pulse runs it through the same five steps every time. What it learns feeds Ask, People, and the agent drafts alike. One source of truth, many views on top.

  • No duplicates. Pulse can re-run any step without creating duplicate entries.
  • Audited. Every external read or write is logged, and so is every ACL decision.
  • Honest. Confidence numbers re-tune weekly from your team’s thumbs-up and thumbs-down feedback.
  1. 01connect

    Connect your tools.

    Read-only access to Slack, GitHub, Notion, Linear, Calendar, Drive, Confluence, Jira, Meetings.

  2. 02protect

    ACLs replicate.

    Channel privacy, doc sharing, and repo permissions all carry over, so Pulse shows each person only what they can already see. Refreshed hourly.

  3. 03read

    Pulse reads the work.

    It picks out the decisions, commitments, features, and past failures from everyday activity.

  4. 04answer

    Hybrid retrieval + ACL gate.

    Pulse weighs meaning, keywords, and recency, favors what your team relies on, and never returns anything you are not allowed to see.

  5. 05learn

    Learns from feedback.

    Your team's thumbs-up and thumbs-down tune the confidence score, so 84% really means right about 84% of the time.

anatomy / 08

Wednesday, 100-person company. No one is checking five tools.

Pulse is what they touch. The five tools sit underneath. Scroll through one ordinary Wednesday at a 100-person company.

  1. Home · 08:55

    Opens Pulse before the day starts.

    Three overnight changes plus a 90-second brief for her 9:30 1:1 with Therese.

    Clarissa · Eng Manager
  2. Ask · 10:15

    Hits a Postgres connection-pool problem.

    73% confident answer in 2.4s: two Slack threads, a fix PR, three ranked experts. Pings Yusuf, who already has the context.

    Therese · IC
  3. Map · 12:30

    Plans a CSV import feature.

    Map → Planner tab pulls four similar past features with real estimates and pre-built risks. Forty minutes saved.

    Adekunle · PM
  4. Ask · 15:00

    Prepares the board update.

    Deep-research mode summarises the quarter (features, blockers, velocity) in 90 seconds. Four hours saved.

    Apoorv · CTO
  5. Actions · 16:15

    Clears the agent-action drafts inbox.

    Three drafts: a stuck-PR Slack nudge, a 1:1 reschedule, a missing-rationale Linear comment. Tweaks one tone, approves all.

    Clarissa · Eng Manager
  6. Home · 17:30

    Wraps up the day.

    Home flags a promise due, a #design question he likely knows, a PR waiting 18h. Sends, answers, pings. Done.

    Kabir · Designer
08:55 → 17:30
scroll →
capability surface / 09

One platform. Most capabilities you'll meet on the way to something else.

A flat list would be useless. Here's the structural shape of the platform; every feature lives somewhere on this map.

Decision memory

Auto-captured with rationale, easy to revisit.

Replay the alternative

What would the other choice have looked like?

Failure case library

Patterns this team has been burned by.

Pre-mortem generator

3-5 plausible failure modes per draft.

Hybrid retrieval

Vector + lexical + recency + behaviour.

Counter-evidence mode

Argue the strongest case against your own answer.

Ask about the past

"As of 30 days ago" for retrospectives.

Brainstorm mode

Divergent. No verdict, no confidence score.

Agent action inbox

Drafts everywhere, send only here.

Five-minute undo

Retract anything Pulse sent for you.

Playbooks

3-5 actions bundled into one click.

Send limits

Approved domains, channels, and teams only.

Stuck detection

Items on you over 18h surface gently.

Champion watch

Notifies when a customer's champion leaves.

Catches stuck work

Stalled stages nudge their owners.

Decisions that stalled

Decided but unbuilt after 60 days.

Honest confidence

Accuracy tracked per team and per topic.

Skills compiler

Exports portable Agent Skills.

MCP server

Query Pulse from Claude / Cursor.

BYOK

Bring your own Anthropic key.

Custom entity types

Model your own kinds: Vendor, Incident, Experiment.

Drop-anywhere capture

An iOS Shortcut or a watched folder, into your inbox.

Self-tuning skills

Pulse proposes sharper instructions; you approve.

the payoff

We make your newest hire as smart as your most senior.

Pulse hands every person the decisions, playbooks, and experts the company already has. The whole team levels up together, and the company keeps getting smarter, not just your AI.

Your unfair advantage. Like giving your team a computer when everyone else is still on paper.

manifesto / 10

Things Pulse will never do.

Your company brain is dangerous in the wrong shape. We've drawn the line, in writing, on the marketing page. Hold us to it.

privacy & power · never
  • No individual productivity scores. No scores, no rankings, no “underperformer” detection. Work context on a person's page is visible to everyone equally, including that person.
  • No manager-only surveillance. Managers see memory aids, not rankings. Sentiment is team-level only, with a five-sender floor.
  • No training on your data. Customer content stays in the customer's tenant. Not aggregated, not anonymized-for-fine-tuning.
  • No permission expansion. If you can't see it in Slack, you can't see it in Pulse. Period.
trust & control · never
  • No silent auto-execution. Every external write is approved by a human or auto-approved by an explicit policy you wrote.
  • No vendor lock-in. Export your whole workspace as JSON whenever you want. Playbooks export as portable files.
  • No notification firehose. Alerts are tuned. Quiet by default, loud only when it matters.
  • No fake confidence. Every confidence score is re-tuned weekly against your team's own thumbs-up/down feedback, per workspace and per topic, not against a generic benchmark.
questions / 07

The questions every founder asks before they trust us with their tools.

Ask anything else at support@pulsehq.tech.

How is this different from enterprise search?+

Enterprise search indexes pages. Pulse does three things a search box can't. First, it remembers decisions: each one is saved with the reasoning, the owners, and the sources behind it, so you go straight to it instead of re-Googling. Second, it drafts the work that should happen next (Linear tickets, Slack DMs, calendar invites) and lands it in your inbox to approve. Third, every answer shows where each line came from and how sure it is. The chat box is the easy part; the memory and the honest confidence are the hard part.

Can my coworker see things I shouldn't see?+

No. Pulse copies the permissions you already have. Slack channel privacy, Confluence page restrictions, GitHub repo permissions, and Drive file ACLs replicate during sync. The one caveat is Notion: its API exposes no per-page permissions, so Notion content is scoped to the installing admin or the whole workspace, a choice made at connect time. Every answer, search, briefing, and draft is filtered to what you are allowed to see before it reaches you. When Pulse spots something you don't have access to, it tells you and offers a one-click request that goes back to the source tool. It never leaks snippets, and access requests, agent reads, and API calls are written to the audit log.

What happens when an agent action goes wrong?+

You have five minutes to undo any action Pulse sends. The undo retracts the message, comment, ticket, or page in the source tool. Before anything sends, Pulse checks it against your limits (which email domains, channels, repos, and teams are allowed) and rate caps, so a mistake can't fan out. After five minutes the log entry is permanent, but you can still reverse the action by hand.

How accurate is the AI?+

Pulse learns how accurate to be from your own team. When it says 84% confident, it is right about 84% of the time on that kind of question in your workspace, and that number re-tunes from your thumbs-up and thumbs-down each week. Every answer marks which parts came from a real source and which are inferred, so you always know how much to trust it.

Will you train models on our data?+

No. It's in the contract, on the security page, and built into how Pulse works. We call Claude and OpenAI under API terms that bar training on our traffic; the providers may retain inputs briefly (up to 30 days) for abuse monitoring. Your data stays in your own isolated space, and nothing crosses between companies for model training. The optional benchmark network only ever shares anonymized ranges, never your raw data.

Can we self-host or BYO key?+

BYOK Anthropic ships today on the Enterprise plan. Paste your own key from /app/admin/byok and every Claude call routes through it, encrypted at rest and decrypted only when needed. AWS Bedrock and Azure OpenAI are available as enterprise add-ons; email enterprise@pulsehq.tech to set them up. Full self-hosting isn't offered yet; it's on the longer-term roadmap for regulated industries.

How long does setup take?+

Three phases. (1) ~30 minutes of conversational interview with a founder (Slack-style; we ask about your tools, decision rituals, who owns what) so the graph isn't cold on day one. (2) ~10 minutes of OAuth handshakes: Slack, GitHub, Notion, Linear, Calendar; pick the connectors that match your stack. (3) 2 to 6 hours of background backfill before retrieval is dense. You'll get useful answers within minutes for recent activity, but the long-tail of historical decisions takes the rest of the first day to land.

Every decision remembered,
every reason a click away.

Give your team the institutional memory of a ten-year company on day one. We're onboarding teams privately, book a 20-minute demo and we'll show Pulse running on a workspace like yours.