AI Agents

Tell your agent what you need in plain language, and the Miravo Skill lets it find, build, run and check the factory for you. You don't need to learn the YAML schema or the CLI.

The skill works with Claude Code, Codex and any agent that loads Agent Skills. An agent without skill support can still drive Miravo through the CLI if you point it at the skill's SKILL.md.

InstallCopy Anchor Link

The agent needs the miravo CLI on PATH; see Install Miravo. The install script offers to install the skill for you. To do it yourself, or to add it later:

miravo skill install

This writes the skill (a SKILL.md, a references/ folder and agents/openai.yaml) to ~/.claude/skills/miravo for Claude Code and ~/.agents/skills/miravo for Codex and other Agent Skills tools. The files are built into the binary, so no clone is needed.

Rerun it after upgrading Miravo to update the skill (add --force if you edited the installed copy). The install script reads MIRAVO_INSTALL_SKILL: 1 installs without asking, 0 skips.

Install manuallyCopy Anchor Link

The skill is the skills/miravo folder of the Miravo repository. Link or copy it yourself if you prefer to track the repository.

# all projects
mkdir -p ~/.claude/skills
ln -s /path/to/miravo/skills/miravo ~/.claude/skills/miravo

# one project: copy the folder instead
mkdir -p .claude/skills
cp -R /path/to/miravo/skills/miravo .claude/skills/miravo

The skill is tested by running real requests with Claude and GPT-based agents against the released binary.

Use itCopy Anchor Link

Ask the way you would ask a colleague. You don't need to know the schema or write a long prompt.

• "Give me a bottling line I can point Grafana at over MQTT, and show me what happens when the capper jams."
• "I need OPC UA data from a water plant for my historian."
• "Make a compressor that wears out over a simulated month so I can test my predictive-maintenance app."
• "My SCADA polls Modbus: simulate a building's HVAC and give me the register map."
• "Export the backup power facility for Litmus Edge."
• "Give me a repeatable run for my CI tests."

What the agent doesCopy Anchor Link

1. Checks the built-in catalog for a template or model that already fits.
2. Reuses it, or writes YAML only for what the catalog lacks.
3. Validates the files.
4. Runs the factory on free ports.
5. Checks that values actually change.
6. Injects faults if you asked for them.
7. Reports back.
8. Stops the run and cleans up.

At the end it tells you how to connect (endpoints, topics, node IDs or registers), what it assumed, and the evidence that the data is moving.

What it can doCopy Anchor Link

• Build lines and whole plants.
• Write custom equipment when the catalog has none.
• Add wear and faults, including faults that affect other assets.
• Change a running factory.
• Produce the protocol map for MQTT, OPC UA and Modbus.
• Export for Litmus Edge.
• Make runs reproducible.

Related pagesCopy Anchor Link