eisy meets AI!
The future of eisy is here, and it changes everything.
Meet eisyAI. Command and control devices and plugins, create and analyze complex programs, and even get help diagnosing issues, all in plain English.
SEE IT LIVE: OpenADR Webinar
Join us for a live demonstration of eisyAI at the OpenADR webinar where we’ll show you what’s possible:
- Command/control/monitoring of one or many devices using English as if you are talking to a human
- Creation of complex conditional logic programs right in front of your eyes
- Real capabilities, real devices, real scenarios
- Live Q&A
📅 Date: Thursday, January 29th 8AM US Pacific Time
🔗 Register now: [ link ].
[ link ].
WHAT’S NEXT
Soon after, we’ll have another webinar going deep into what you can do with YOUR own eisy and your configuration. Stay tuned for details.
January 20, 2026 @ 7:35 am
Please post the webinar on YouTube so more of us can watch the information about the ai integration. Thank you
January 20, 2026 @ 7:45 am
Hi Robert,
More than a post, we’ll have a webinar before our launch (end of February) with everything you need to know including installation and use cases. Please stay tuned.
January 20, 2026 @ 9:23 am
Is this webinar openADR specific, or covering other uses of EISY?
January 20, 2026 @ 9:34 am
Hi Ed,
It covers all concepts + creates IoX programs at runtime. Some of the use cases are related to OpenADR price signals and some are not.
January 20, 2026 @ 11:54 am
This is fantastic! I built my own MCP server for ISY and have it working as a Steamable HTTP MCP server endpoint for n8n workflows, as well as Claude Desktop Stdio tool library. The most interesting aspect of the project wasn’t the MCP or LLM API interaction- it’s all the LLM context caching and layers of tool instructions, especially next-step/toolchain continuity thought multiple tool calling prompt threads.
Can’t wait for the webinar, thanks for setting this up. AI chat is now the primary interface I use to manage devices in my house, and it will be great to hear what others are doing.
(Example conversation below)
“summarize the devices and status in the den. Group by function and related controls. Advise on inconsistent configuration or data which could improve your interaction and tool token efficiency.”
Den Device Summary & Status
Lighting Devices (3 Total)
1. Den Main (Ceiling Light)
Address: XX 13 5B 1
Type: DimmerLampSwitch_ADV (Standard dimmer)
Current Status: OFF (0%)
On Level: 70%
Spoken Name: “den celing” [sic]
Location: Not set
2. Den Couch Lamp
Address: XX D3 20 1
Type: DimmerLampSwitch_ADV (Standard dimmer)
Current Status: ON at 80%
On Level: 100%
Spoken Name: Not set
Location: Not set
3. Den Ollie Lamp
Address: XX A 89 1
Type: DimmerLampSwitch_ADV (Standard dimmer)
Current Status: ON at 80%
On Level: 100%
Spoken Name: Not set
Location: Not set
Configuration Issues & Recommendations
Critical Issues:
Missing Location Field ⚠️
Problem: None of the 3 den devices have their location field populated
Impact: The find_nodes_by_location() tool cannot discover these devices, requiring name-based searches instead
Fix: Set location = “den” for all three devices in ISY admin console
Token Efficiency: This would enable location-based queries and reduce search complexity
Inconsistent Naming Convention ⚠️
Problem: Device names use different formats:
“Den Main” (space, proper case)
“Den couch lamp” (space, mixed case)
“Den Ollie lamp” (space, mixed case, personal name)
Recommendation: Standardize to “Den – [Function]” format:
“Den – Ceiling”
“Den – Couch Lamp”
“Den – Ollie Lamp”
Missing/Incomplete Spoken Names ⚠️
Problem:
“Den Main” has spoken name “den celing” (typo)
Two devices have no spoken names at all
Impact: Voice control inconsistency, AI interpretation difficulty
Recommended Spoken Names:
“den ceiling” (fix typo)
“den couch lamp”
“den ollie lamp”
No Functional Grouping ⚠️
Problem: Cannot identify which lights serve which purpose from device data alone
Recommendation: Add descriptions:
Den Main: “Overhead ceiling light”
Den Couch Lamp: “Table lamp near couch”
Den Ollie Lamp: “Ollie’s table lamp” (or provide context for who/what Ollie is)
Token Efficiency Impact:
Current Workflow (inefficient):
Cannot use find_nodes_by_location(“den”) → fails
Must manually search context reference for “Den” in names → 220 rows to scan
Must call individual device status or batch calls with hardcoded addresses
With Proper Location Field (efficient):
Single find_nodes_by_location(“den”, include_status=True) → instant results
Reduced token usage by ~60% for typical den queries
Enables room-based automation discovery
Additional Observations:
Consistent behavior: All three dimmers have same ramp rate (0.1 seconds)
Different manufacturers: Den Main is type 1.32.69.0, while both lamps are 1.14.67.0
No scenes detected: No multi-device scenes involving these lights were immediately apparent
Immediate Action Items (in order of impact):
Set location = “den” for addresses: XX 13 5B 1, XX D3 20 1, XX A 89 1
Fix typo: spoken name “den celing” → “den ceiling”
Add spoken names for the two lamp devices
Add descriptive text to clarify “Ollie” context
After changes, run refresh_cache() to update the system
January 20, 2026 @ 4:37 pm
I mentioned this years ago, but why isn’t there a Random Function available, so that the time for sunset can be intelligently randomized, or lights randomly going on/off to simulate someone being home (e.g.vacation mode)?
A way of removing static/ predictable lighting is a real plus for security
January 20, 2026 @ 4:50 pm
Maybe I missed it, but where is this program edit screen in the GUI? I can see my node, my existing programs and everything, but nowhere to enter edit mode. Or is editing programs not yet allowed in the GUI? For that matter, when does eisyAI become available to EISY users, Jan 29th or earlier? Looking forward to this huge jump into the present world of AI.
January 20, 2026 @ 7:06 pm
Because you can do it easily with a Wait (Random).
January 20, 2026 @ 7:11 pm
Hi Patrick, this is amazing! This said, you can optimize all this using the nucore-ai library. Just clone the repo and read the README. Under the Assistant, there are Claude, OpenAI, as well as Generic (runs on local hardware with GPU using llama.cpp) implementations. It serialized nodes with their nodedefs and incorporates it into a prompt.
Since it’s all open source, it would be awesome to also include you MCP server.
January 20, 2026 @ 7:12 pm
Hi Rod,
Edit is coming shortly. For now, it’s just view, rename, and remove.