Claude Code repairs Home Assistant presence triggers
Adam Beguelin used Claude Code to inspect Home Assistant traces, identify repeated home-to-home camera triggers and deploy transition filters through a versioned configuration workflow.
blog.beguelin.com
Original prompt
Shared by the source author. See source notes for context.
why did the automation dresser camera off when home just run?
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source author
- Adam Beguelin
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- Beguelin reports a deployed filter correction after Claude traced five redundant camera-off runs to home-to-home state updates.
- Human involvement
- Beguelin reports the symptom, approves the proposed fix and reviews changes; long-term absence of repeated triggers is not independently measured.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- Ask why the camera automation fired and retrieve its execution traces.
- Identify home-to-home state updates as the repeated trigger.
- Approve explicit arrival and departure filters in the YAML.
Adam Beguelin noticed that a Home Assistant automation for turning off a camera when he arrived home kept firing. He asked Claude Code why it had just run. Through an MCP connection, the agent read execution traces and found repeated home-to-home updates rather than real arrivals.
The trigger listened to any person-state update, so location refreshes repeatedly switched off an already-off device. Beguelin reports five runs in forty minutes. Claude proposed explicit arrival and departure transition filters, and Beguelin approved the change.
The agent edited the YAML locally, copied it to Home Assistant, reloaded automations and committed the change. The broader described workflow checks whether the configured entity exists and is enabled after deployment. This supports an applied configuration fix, but not a long-term independent measurement that the unwanted triggers never returned.
Other examples in the account include changing bedroom brightness and designing motion lights that preserve lights already on. Those examples establish the surrounding maintenance context rather than separate outcomes here. Beguelin retained the conversation, reviewed proposals and supplied course corrections. The useful result is tracing an observed household problem to its actual event condition before changing and deploying the configuration.
Prompt
Original prompt
Source: Blog · Adam Beguelin
Frequently asked questions
What caused the repeated runs?
The automation reacted to every person-state update, including home-to-home location refreshes.
What was changed?
Explicit from/to filters limited the triggers to actual arrival and departure transitions.
Who approved deployment?
Beguelin told Claude to make the proposed change before it edited, deployed and reloaded the configuration.
More details
Remaining steps
- Deploy, reload, verify configuration state and commit the change.
About this account
Claude Code reads HA traces, identifies the repeated transition, edits YAML, deploys/reloads it and commits the change.
Firsthand primary account opened and read; original publication date checked in primary metadata or visible timestamp. Outcomes are author-reported, not independently reproduced. Applied configuration fix; no independent longitudinal trigger audit or household safety guarantee. Original prompt excerpt reproduced exactly from the source; excerpt only, not the complete prompt.
Credit: Adam Beguelin
Source published
Added to Agents In Use
Source checked
Tools and topics
Claude Code · Home Assistant · MCP server · YAML · scp · Git
Life admin · Reported deployment
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.