Pulse | Miragon AI

The problem

Every question about a running process waits for a developer

When a process gets stuck, the people who hear about it first, such as operations staff, support or a product owner, usually cannot see why. They ask a developer, so every question waits for the few people who know each engine's Cockpit and PromQL.

Today

One Cockpit per engine
Stack traces, job by job
PromQL for runtime questions
Tickets by copy and paste

With Pulse

One question covers one or several engines
Incidents grouped by failing activity and type, down to the trace
Activities ranked by execution time and incident rate, BPMN heatmap
A ticket draft with title, body and labels

How it works

Ask in the chat, get the engine's answer as a widget

The runtime figures need CIB seven with the metrics plugin. Everything else works on Camunda 7, CIB seven and Operaton.

Connect
Pulse reads your engines over their REST API, never the engine database, and needs no migration.
Ask
You ask in Claude, ChatGPT or another MCP client, and the assistant calls Pulse's tools.
See
The cockpit view, incident clusters, BPMN diagrams and dashboards come back as widgets in the chat.
Decide
Pulse drafts the ticket with title, body and labels. Where it goes is your decision.

Try it

Connect your Claude in 30 seconds

Get started

Live playground

Talk to your processes.

No install, no login. Open Pulse's hosted playground in an MCP chat, connected to a live, seeded CIB seven engine. Ask a question and watch the interactive widgets render.

Try it out

Try askingShow me the Cockpit

Safety

Read-only until your team decides otherwise

The AI collects the state of your engines and proposes the next step. Whether it may change anything depends on the toolset the server runs with, and widgets hide the buttons for actions the toolset does not allow.

read-onlyThe default without OAuth and on the playground. Queries, widgets and ticket drafts.
operationsThe default behind OAuth. Adds starting instances, completing tasks, retries and resolving incidents.
adminOnly when an operator names it. Adds deletions, modifications, suspensions, migrations and signals.

Deployments need admin plus CAMUNDA_ALLOW_DEPLOYMENTS=true. Because the default cannot change an engine, you can hand Pulse to a wider group, such as support or product owners.

Get started

Run it on your own engine

You need Docker and the REST API of a Camunda 7, CIB seven or Operaton engine. The Pulse docs have the command.

Get started

Status

Where Pulse stands on 14 October 2026

Pulse is at v0.20.0 and pre-1.0. The last three minor releases each list breaking changes, such as renamed tools and result fields, so pin the image tag you run. The public playground runs the same version.

Drafts, never files
Pulse drafts tickets and never files one.
Runtime figures on CIB seven
They need the metrics plugin and cover only traffic recorded after it was installed, at most 30 days back.
No process mining
The analytics module knows neither waiting times nor variants, so it is no tool for Optimize.
Camunda 7, CIB seven, Operaton
Pulse speaks their REST API. Camunda 8 is out of scope.

Proof you can open

Everything here is public

Where it fits

Pulse in the lifecycle

Docs

The full reference is in the docs

The README is the full reference for every tool, toolset and environment variable.

Run Pulse on your own engines.

Tell us which engines you run and who asks about them today. You get an honest assessment of whether Pulse helps first.

Talk to us

A 30-minute first call.

PulseAsk your engines what is stuck and why.

Connect

Ask

See

Decide

Talk to your processes.

Drafts, never files

Runtime figures on CIB seven

No process mining

Camunda 7, CIB seven, Operaton

Run Pulse on your own engines.