Alrakis
Alrakis

Run an incident from the chat you already use.

Alrakis is an incident response bot for Slack, Microsoft Teams, and Lark with one engine behind all three. Declaring an incident opens a room, assigns roles, starts the bridge, pages on-call, opens a ticket, and keeps the rooms, the status page, and the record in step until the postmortem.

Below is the real engine, running in your browser against an in-memory workspace. Type commands as any of the four people, or let the scripted incident play.

Four people share this workspace: Alice, Bob, Carol, and Dana. Nothing leaves your browser; reload the page and it is gone. The bridge is a stand-in link, and no pager, ticket, or status page provider is connected here.

What happens when someone types inc declare

chat platform  ──▶  adapter  ──▶  router  ──▶  engine (guards)  ──▶  event log (append-only)
(Slack / Teams /     parse         grammar        commands                 │
 Lark)                                               │                     ▼
                         rooms ◀── fan-out ◀─────────┤              projections (incidents,
                    (every platform)                 ▼              roles, tasks, updates...)
                                            integration plugins
                              (bridge, pager, ticket, status page, email, webhooks, yours)

One incident, every platform

An incident can have rooms on Slack, Teams, and Lark at once; every room sees the same cards, and people are matched across platforms by email.

The playbook carries the process

Severities, roles, checklists, custom fields, workflows that act on events, reminders with conditions, escalation targets, response targets, a service catalog, and the postmortem's layout, in one YAML file per incident type.

The record writes itself

Key times, leads and what ruled them out, decisions, promised updates and whether they were kept, pinned messages with their images, and a chat transcript with its SHA-256 at the close.

Integrations as plugins

PagerDuty, Opsgenie, Jira, ServiceNow, Linear, GitHub, Confluence, Notion, Statuspage, Zoom, Meet, email, Twilio, webhooks, and a Claude assistant for drafts; or a plugin file of your own, loaded from the config.

Append-only, replayable

Every change is an event. The state is a fold over the log, the same in memory and in Postgres, and a parity suite keeps the two stores agreeing.

Measured

Reports give counts, time to resolve, time to detect and to mitigate, targets met, and when incidents happen; the timeline streams live to a terminal or a screen on the wall.

Run it yourself

From npm

npm install -g alrakis
alrakis init          # asks about your platform and integrations
alrakis check --probe # every credential, against the vendor
alrakis serve

Or Docker

docker compose up -d   # alrakis + Postgres
# Kubernetes: a Helm chart on ghcr.io, or kustomize
# systemd: one bundle file, no node_modules

A scripted incident with no accounts, like the one above, runs in the terminal with alrakis demo. The docs cover getting started, configuration, playbooks, the CLI, the API, alert receivers, plugins, and the Slack, Teams, and Lark setups.