Your workspace, as a live process graph.
Pulse connects to your tools, keeps track of the decisions and commitments buried in everyday work, and answers anything you ask with the source for every line. Every stage is permission-aware. Here is how, step by step.
Six steps, from your tools to an answer you can act on.
Connect your tools
step_01 · connectPulse reads from each tool you connect, read-only. Your data stays walled off from every other company’s, encrypted, and you can bring your own AI key on Enterprise.
// per-tenant ingest lane (read-only) stream slack.events → "pulsehq.ingest.slack" // 4.2k/min stream github.webhooks → "pulsehq.ingest.github" // 91/min stream notion.delta → "pulsehq.ingest.notion" // 38/min stream linear.events → "pulsehq.ingest.linear" // 17/min stream calendar.changes → "pulsehq.ingest.cal" // 6/min // retention: 30d hot · 365d cold (configurable) // rate limits: respected upstream · backoff on 429
- ▸Read-only, every tool. Pulse never writes anything back while it reads. Sending actions uses a separate, limited permission.
- ▸ACL captured at the source event. A Slack message’s channel ACL travels with it through every downstream stage.
- ▸Reconnect safely. Disconnect and reconnect a tool and Pulse picks up where it left off, with no duplicates.
Five choices no other tool makes, side by side.
Most company search just indexes documents and asks you to trust the vendor. Pulse takes a different path on each of these.
Built on permission, not against it.
If you can’t see it in Slack, you can’t see it in Pulse. Three structural guarantees, not three policies.
Source-system ACLs, mirrored.
Pulse never widens access. A Slack DM you couldn't see stays invisible. Pulse can tell you it exists and offer a request-access flow, but it never shows the contents.
Mark something confidential, and it stays that way.
Flag a thread or doc as confidential. Pulse can help you think it through but won't remember it, surface it, or show it to anyone else.
“Forget this” is a one-click action.
Mark anything forget and it's removed from Pulse. There's an auditable record that it was forgotten; it's gone from answers and from search.
Everyone can see what Pulse knows about them.
A built-in personal-data page (Settings → Data) shows everything Pulse has picked up about you. Export it. Delete the parts that aren't shared.
Your data is walled off from every other company's.
One company can never see another's. Storage and keys are separated, and BYOK is available on Enterprise.
Every answer and every action is logged.
Set your own retention window. Export the log as CSV or JSON today; live streaming to your security tools is on the Enterprise roadmap.
Four choices that shape everything Pulse does.
Pulse looks the way it does because of these four beliefs. Each one shapes a hundred smaller decisions inside the product.
Permission-aware, never expanding.
Every answer, search, briefing, and draft is filtered to what you are allowed to see before it reaches you. We mirror Slack, GitHub, and Notion ACLs into the graph at sync time and enforce them at retrieval time, and there is no path that bypasses this filter.
Process graph, not doc graph.
Decisions, commitments, features, and past failures are saved as things you can look up, not pages re-read every time you search. The memory is the hard part; the chat box on top is the easy part.
Memory ages on a curve.
Different things go stale at different speeds. Pricing ages in 30 days; mission lasts 5 years. Pulse leans on recent sources and flags an answer when it is leaning on something old.
Calm by default.
Pulse never paints individuals as the problem. It flags stuck stages, not stuck people. Briefings are role-specific, not all-staff. Actions wait for your nod and undo on demand. The product is opinionated about restraint.
From a single draft to a multi-step job, on your terms.
Custom agents go beyond one prompt, one action. An agent can think, take a step, look at the result, and keep going until the job is done or it hits the limits you set. And the limits are real: you decide how far each agent can go, what it can see, and how many steps it gets before anything runs.
Every step, recorded and replayable
Every step an agent takes is recorded: what it thought, what it did, what it saw. Admins can replay any run end to end, with the full prompt, every action, and the timing.
Suggest, draft, or execute, per agent
Set how much rope each agent gets. Suggest just shows the idea. Draft fills your approval queue but waits for a human. Execute sends after the same five-minute undo window. Change it any time; the next run picks it up.
Agents can hand off to other agents
One agent can call another. A 'research the customer' agent can feed a 'draft the renewal email' agent, and the result flows back so the first can decide what to do next. The same limits and permissions apply to every agent in the chain.
Public forms that kick off an agent
Build a simple form (text, dropdowns, file upload). A submission kicks off the agent with what was entered. Handy for help-desk intake, qualifying sales leads, anything that needs a trigger from outside the company.
Other systems can trigger an agent
A failed Stripe payment, a successful GitHub deploy, a closed Linear cycle, each can kick off an agent run. Every incoming request is cryptographically verified before Pulse trusts it.
Tested before it is trusted
The reusable skills agents run are vetted before they're trusted: each flows DRAFT → SHADOW → ACTIVE → DRIFTED → ARCHIVED. In shadow mode a skill observes humans on similar tasks and only graduates to active once its match-rate clears the bar; if its results later drift it flips to DRIFTED and pauses for review.
Every agent follows the same safety rules: personal data stripped from anything it sends out, reads limited to what its creator can see, a paid plan required to execute, and five minutes to undo any external action. Autonomy without surveillance, by design.
From signup to first agent action: about 30 minutes.
Fast enough to roll out mid-sprint, at 5 people or 500. The 30-minute guided interview is what makes Pulse work for any team.
Connect Slack + GitHub
Connect each tool. Pulse starts reading immediately, read-only, and never writes anything back.
Guided interview
30-min conversational onboarding with a founder. Pulse learns your vocabulary, processes, and pain points.
First decisions extracted
Pulse surfaces 8 to 12 decisions from the last 90 days to confirm. Edit freely.
First drafted action
Pulse runs overnight. The first drafted nudge or ticket lands in your approval inbox.
From the blog
Process graph vs document graph
Why every enterprise AI search tool indexes documents, why that flattening loses the structure that makes team knowledge valuable, and what a process graph unlocks instead.
Read →Auto extracted Skills from observed work
Human authored agent workflows go stale the moment they are written. Auto extracted Skills stay current because they reflect what teams actually do, and they travel across every Skills compatible AI tool.
Read →Five questions your AI tool should answer with sources
Sentence level attribution is the difference between a verifiable answer and a plausible one. Five concrete questions to ask any AI tool, and the architectural reason most tools cannot deliver it.
Read →30 minutes, not 30 days.
Bring three connectors and a half-hour conversation. Pulse will have a working model of your team before lunch. Or walk the demo workspace right now, no signup.