One brain. Every channel.
Post where you already talk, ship where you already ship. The tags say exactly how each integration is wired today, and the AI is the same whichever channel the message came through.
Chat
- K9 Chat: native, end-to-end encrypted; the assistant reads check-ins and direct messages, and tells the sender when a message could not be decrypted
- Slack: bridged, multiple workspaces; check-ins, leave approvals and "/pulse me" work from Slack
- Microsoft Teams: bridged, with a per-tenant app manifest
- WhatsApp: outbound on-call reach for escalations
- Google Calendar and Microsoft 365: through n8n
Work
- Jira: reads assigned issues and maps their status into the scorecard
- GitHub and GitLab: activity, used to verify the work named in check-ins
- VPSie: native; the assistant lists, inspects and manages servers in a rolling conversation
- Bitbucket: soon
Automation and events out
- n8n: native, both as an AI provider (your own pipeline answers) and as an automation target
- Outbound webhooks: signed JSON to any HTTP endpoint, retried for hours, paused after repeated failure, with a test button; reach Splunk, Datadog, Grafana, PagerDuty or Zapier this way
- Email: SendGrid or any SMTP server for reports, digests and escalations
- REST API with personal tokens; CSV exports that a BI tool can pull on a schedule
AI providers
One interface, 16 providers. The admin sets the active provider and an ordered failover chain; if one is out of credit or rate-limited, the next answers, and each Ask AI reply is tagged with the provider that produced it.
| Local | Ollama |
|---|---|
| Your pipeline | n8n webhook |
| Paid | Anthropic, OpenAI, DeepSeek, Cohere |
| Free tier | Google Gemini, Mistral, Groq, OpenRouter, Perplexity, Together, Cerebras, Hyperbolic, Hugging Face, Cloudflare Workers AI |
What each tag means
| native | Built into the product; no bridge, no third-party relay |
|---|---|
| bridged | Reached through a chat bridge; check-ins and replies work, presence comes from the bridged client |
| native · E2EE | End-to-end encrypted on your own Matrix server; InRoots reads the room like any member |
| activity | InRoots reads events to verify check-ins; it does not write back |
| via n8n | Reached through your own n8n flow, not a built-in connector |
| events out · webhooks | InRoots emits signed events you route anywhere |
| Outbound delivery through SendGrid or your own SMTP server | |
| on-call reach | Outbound messages only |
| local | Runs on your host; nothing leaves the network |
| soon | Planned, not shipped; not counted in any number on this site |
How the AI uses them
The work integrations are what make a check-in verifiable: a task named in the morning is matched against the day's Jira, GitHub and GitLab activity before it earns Progress credit. The chat integrations are where coaching, summaries, leave decisions and blocker routing are delivered, in the same thread the person already uses. The AI provider is the only piece that thinks; everything else moves evidence.
Adding one
Integrations are configured in the admin settings. Credentials are encrypted at rest with a key you hold, every connection has a test action, and a failing endpoint is retried and then paused rather than silently dropped. Availability of Slack, Teams, Jira and white-label depends on your plan.