Talk to us

Miragon AILifecycle

ExecuteFind what is stuck on your engines, and why.

Your team finds out what is stuck on the running engines and why, and takes the next safe step. Both tools in this phase work with Camunda 7, CIB seven and Operaton.

See the tools
Talk to us

The problem today

What hangs is spread across Cockpit tabs and stack traces

Today

  • To see what hangs, someone clicks through the Cockpit tabs of each engine and reads stack traces job by job
  • Runtime questions such as how long instances take usually need PromQL against Prometheus
  • Incident tickets start as copy and paste into the tracker
  • A coding agent takes the same path through the raw REST API, one request at a time

With Pulse and operate

  • Pulse answers in the chat with widgets, such as incident clusters by failing activity and incident type, down to the stack trace
  • Runtime figures such as durations and incident rates need no PromQL (CIB seven with the metrics plugin, 30 days at most)
  • For an incident, Pulse drafts a ticket, and you decide where it goes
  • operate status triages the whole engine in one read-only call and names the next command for each finding

How Miragon AI helps

Pulse in chat, operate in the shell

Both tools read the same engines through their REST API. Pulse is a self-hosted MCP server (the repo Miragon/miragon-ai) that your team runs and asks in chat. operate is a CLI without a server for developers, coding agents and CI.

  • Process operations in chat

    Pulse

    Your team hosts Pulse once, and operations, support and product owners ask it about running processes in Claude, ChatGPT or another MCP client. It reads one or several engines without a migration or access to the engine database, and answers with widgets in the chat, such as incident clusters by failing activity and incident type, down to the stack trace. Without OAuth it starts read-only. For an incident it drafts a ticket, and you decide where it goes.

    The Pulse page links the public playground, where you can ask it first.

    Statusv0.20.0 (released 2026-10-13), pre-1.0, and each of the last three minor releases brought breaking changes. Runtime figures such as durations and incident rates need no PromQL, but they work only on CIB seven with the metrics plugin and reach back 30 days at most.

    More on PulseGet started

  • Engine state for shell, agent, CI

    operate

    operate turns the 396 operations of the engine REST API into commands for your shell, a coding agent such as Claude Code, or a CI job. operate status triages the whole engine in one read-only call and names the next command for each finding, and operate inspect --business-key ORD-1001 shows where one instance waits and why.

    With --fail-on warning, any finding (an open incident included) fails the CI job.

    Status0.7.0 (released 2026-10-14) on npm, as standalone binaries for Linux, macOS and Windows and as a container image. Pre-1.0, and 0.7.0 brought its first breaking change.

    More on operateGet started

Not Camunda 8 Operate

operate has nothing to do with Operate, the Camunda 8 web app, and it does not support Camunda 8.

Human and AI

The agent reads, your team decides what changes

  • The agentReads the engine state and drafts the ticket.
  • Your teamDecides whether the agent may change anything on the engine (the operations toolset in Pulse, a profile without --read-only in operate) and files the ticket itself.

Hand-over

From a deployed model to tickets and raw figures

Models arrive from Implement. Tickets go back to Implement for the next fix, and raw figures go on to Optimize.

Comes in

  • From Implement

    Models that already run on the engine or that operate deploy uploads, plus the engine's REST URL and credentials. Your own MCP server from the Solution templates runs the way Pulse does. Runtime figures in Pulse also need CIB seven with the metrics plugin io.miragon.mcp:cibseven-history-metrics and Prometheus.

    • .bpmn
    • .dmn
    • .form
    • /engine-rest URL
    • Docker image

Goes out

  • To Implement

    Ticket drafts from Pulse with title, body, labels and an optional prefilled GitHub issue URL. From operate, the root-cause line for the next fix and exit codes for your CI pipeline.

    • Markdown with labels
    • JSON
    • exit codes

  • To Optimize

    Raw material only, since Optimize has no tool yet: Pulse's runtime figures (CIB seven with the metrics plugin, 30 days at most) and operate's raw history queries.

    • JSON and chat widgets (Pulse)
    • JSON (operate)

Find out what is stuck on your own engines.

Tell us which engines you run and who asks about them today. We go through whether Pulse, operate or both fit your team.

Talk to us

A 30-minute first call.