Integrations
How evidence gets into Atisbo — one agent-driven route, four direct ones
Integrations
There are two ways evidence reaches Atisbo, and they answer different questions.
Ask your agent. Around a thousand tools connect this way. You do not go looking for a card in a settings screen — you tell the agent what you want connected, it finds the tool, and it hands you a single authorisation link to approve. This is the route for Zendesk, Intercom, Jira, Linear, HubSpot, Slack and most of what your team already runs.
Send it directly. Four paths belong to Atisbo itself and depend on no third party: forwarding an email, posting to a webhook, dropping in a file, and GitHub. These have their own setup steps because there is something concrete to configure.
Everything Atisbo reads is read-only, enforced in code rather than promised in a policy. A connection that would let Atisbo write into your tools is refused at creation time — see Connect with your agent.
Ask your agent
Use this for anything not in the tables below. There is no waiting list and no "planned integration" — if the tool is in the catalogue, it connects today.
Send it directly
| Path | What it is for | Setup |
|---|---|---|
| Paste or upload | A CSV export, interview notes, a transcript you already have | None |
| Forwarding customer email, support summaries, meeting transcripts | Copy an address | |
| Signal webhook | Anything that can POST JSON — Zapier, Make, n8n, your own scripts | Copy a URL and a secret |
| GitHub | Merged PRs, releases and deployments as operational context | Register a webhook in the repo |
Two more sources are polled rather than pushed, and your agent creates them because they need an API token rather than a consent screen:
| Path | What it is for | Cadence |
|---|---|---|
| Notion | PM notes, research pages, meeting docs | Daily |
| Granola | Meeting summaries, notes and transcripts | Daily |
What happens to anything that arrives
Every route above ends in the same place, so the shape of what you get does not depend on how it got there:
- The text is classified —
problem,idea,question, or leftunclassifiedrather than guessed at. - It is compared against evidence you already have. Related evidence groups into one Opportunity, and a language model confirms every merge before it happens — two pieces of feedback sharing vocabulary is not the same as sharing a cause.
- Evidence that matches nothing yet stays visible as loose evidence rather than being filed somewhere convenient.
- Priority follows from the evidence: how severe it is, how much of it there is, how recent, and how it lines up with the strategy you have recorded.
Pushed sources (email, webhook, GitHub) usually appear within a couple of minutes. Polled sources arrive on their next sync.
Why more than one source is worth the trouble
One customer reporting a problem is an anecdote. The same problem arriving from a support ticket, a sales call and a survey response is a pattern — and Atisbo raises its urgency on its own when that happens, because the volume of evidence and how recent it is are both inputs to the score.
That only works if the sources are genuinely different. Three exports of the same support queue are still one source.
Metric feeds are a separate thing
Numbers do not go through any of the above. Metric providers (Mixpanel, PostHog, Datadog, Amplitude, Google Analytics, or your own REST endpoint) and metric webhooks live under Settings → Sources → Metrics & Analytics and feed the numbers attached to a product area, not the evidence stream. See Webhook API.