A self-hosted AI workspace for chat, issues and docs.
Most workspace AI runs in the vendor's cloud and reads the vendor's copy of your data. Kernyl puts the workspace and the model calls in your own cloud, and checks permissions every time someone asks.
AI is only as private as the place it runs
Workspace AI is most useful when it can read everything: the incident thread, the customer page, the issue history. That is also what makes it hard for a security team to approve. Where are prompts processed? Who keeps the index? Can the assistant show someone a private channel?
Kernyl answers with architecture rather than policy. The workspace, its search index and the model calls all live in your own cloud, and every answer is built only from what the person asking is allowed to open.
Architecture
One workspace, AI included
Chat, Issues, Docs and Calendar in one app. Ask Kernyl answers with numbered sources, summaries condense threads and channels, catch-up shows what you missed, and @Kernyl answers in chat from what the whole channel can see.
What did we decide about changing GET /matches?
Add the nested invoice object and keep invoice_id, so the change is additive and needs no version bump. Nothing is deprecated until v2.1
It's visible to customers, so it goes in the release notes under API.12 Updating the webhook payloads to match is a separate ticket.3
- 1#engineering · Arjun MehtaThread · 15 replies · Sep 29
- 2Release processPage · Engineering space
- 3LED-380 Record accept/reject feedbackIssue · In progress
Self-hosted and vendor-hosted workspace AI
A general comparison. Vendor-hosted tools differ in the details, so check each one's documentation.
Where content lives
- Kernyl
- Your cloud.
- Typical vendor-hosted workspace
- The vendor's cloud, sometimes with a choice of region.
Where AI runs
- Kernyl
- AI models in your own cloud.
- Typical vendor-hosted workspace
- The vendor's infrastructure or model providers they choose.
Permission checks
- Kernyl
- At question time, for every answer.
- Typical vendor-hosted workspace
- Varies by product.
Tools covered
- Kernyl
- Chat, issues, docs and calendar in one app.
- Typical vendor-hosted workspace
- Often one tool per vendor, each with its own AI.
Pricing model
- Kernyl
- Annual subscription by team size; model usage on your cloud bill.
- Typical vendor-hosted workspace
- Per user, often with a per-user AI add-on.
| Topic | Kernyl | Typical vendor-hosted workspace |
|---|---|---|
| Where content lives | Your cloud. | The vendor's cloud, sometimes with a choice of region. |
| Where AI runs | AI models in your own cloud. | The vendor's infrastructure or model providers they choose. |
| Permission checks | At question time, for every answer. | Varies by product. |
| Tools covered | Chat, issues, docs and calendar in one app. | Often one tool per vendor, each with its own AI. |
| Pricing model | Annual subscription by team size; model usage on your cloud bill. | Per user, often with a per-user AI add-on. |
Comparing against a specific tool? See Kernyl as a Slack alternative, a Jira alternative or a Confluence alternative.
FAQ
What does self-hosted mean for Kernyl?
Kernyl is installed in your own cloud, AI included. We don't run a hosted copy and we never hold your data.
Do we need GPUs or our own model hosting?
No. Kernyl uses managed AI models in your cloud. You don't manage model servers.
Which clouds can Kernyl run in?
Tell us where you run, in a public cloud or your own data centre, and we'll confirm the setup on a call.
What does the AI have access to?
Only what the person asking can open, checked at the moment they ask. @Kernyl in a channel uses only what every member of that channel can see.