Skip to content
Honeypod
DE
Request beta

Features · AI and MCP

Your AI assistant asks Honeypod.

Claude, a model in LM Studio or any other assistant with MCP reads your figures and creates drafts if you want. It is off by default. You decide per business what an assistant may do. You see every access in the log.

Request beta All features
On first access Honeypod asks: “Claude wants to read Honeypod AG”. Below it, the note that Anthropic also sees what Claude reads. At the bottom the buttons “Decline” and “Allow”. On first access Honeypod asks: “Claude wants to read Honeypod AG”. Below it, the note that Anthropic also sees what Claude reads. At the bottom the buttons “Decline” and “Allow”.
Settings › AI assistants, switched on for Honeypod AG with the level “Read and drafts”. Below, the assistants: Claude is connected, LM Studio was last connected yesterday. At the bottom, the log with the draft RE-2026-026 by Claude. Settings › AI assistants, switched on for Honeypod AG with the level “Read and drafts”. Below, the assistants: Claude is connected, LM Studio was last connected yesterday. At the bottom, the log with the draft RE-2026-026 by Claude.

Switching on

Off by default, for each business separately.

Under Settings › AI assistants you switch it on for a business and choose what an assistant may do: Read, or create drafts as well. When an assistant connects for the first time, Honeypod asks you whether it may.

Drafts

Drafts yes, anything final never.

With the level “Read and drafts” your assistant creates customers, invoice and quote drafts, notes and follow-ups. It calculates like the editor, language and tax treatment come from the customer. Until you save the draft, it shows who created it. No assistant can send, submit to ELSTER, delete or lock anything.

The draft RE-2026-026 for Praxis am Ahornweg in the editor. At the top the note “Created by Claude · today, 11:14”, below it customer, number, date and the first items. The draft RE-2026-026 for Praxis am Ahornweg in the editor. At the top the note “Created by Claude · today, 11:14”, below it customer, number, date and the first items.

Log

Every access is in the log.

What an assistant has read stays there for thirty days, with the time. What it has created you open from there. “Undo” takes it back as long as nobody has changed it since.

The assistant log: Today Claude read the VAT payable for Q3, the open invoices and the file of Praxis am Ahornweg and created the draft RE-2026-026, next to it “Open” and “Undo”. Yesterday LM Studio read the review. The assistant log: Today Claude read the VAT payable for Q3, the open invoices and the file of Praxis am Ahornweg and created the draft RE-2026-026, next to it “Open” and “Undo”. Yesterday LM Studio read the review.
At the bottom of the sidebar: “Claude connected”, below it “reading open invoices” and on the right the button to disconnect. Below that the running timer. At the bottom of the sidebar: “Claude connected”, below it “reading open invoices” and on the right the button to disconnect. Below that the running timer.

Sidebar

You see who is reading right now.

A connected assistant appears at the bottom of the sidebar. If it is reading, it says what. A click on “Disconnect” ends the connection until you restart the assistant.

  • Claude Desktop
  • Claude Code
  • LM Studio
  • Any with MCP

Assistants

For any assistant that speaks MCP.

MCP is the open standard through which AI assistants talk to programs. Honeypod adds itself to Claude Desktop with one click. For LM Studio it shows the entry with instructions. Any other assistant with MCP gets the same one.

Privacy

0
Ports Honeypod opens on the network for it

Your assistant talks to Honeypod, not to the network.

It reaches Honeypod through a file in Honeypod’s folder that only your user can open. Honeypod itself sends nothing out. What your assistant reads, its provider sees, so with Claude that is Anthropic. With a model in LM Studio everything stays on your computer.

The AI assistants section by default: The box “Allow assistants in Honeypod AG” is unchecked, below it the sentence on privacy. The AI assistants section by default: The box “Allow assistants in Honeypod AG” is unchecked, below it the sentence on privacy.

What’s more

  • Permission per business and per computer
  • No second question after a no
  • Remove assistants one by one or allow them again
  • The question also as a notification when Honeypod is in the background
  • Access only while Honeypod is open
  • Category suggestions for checking in the review
  • Created items also in the audit trail, with their author

Missing something? Then you are probably not the only one.

Suggest a feature

Next

Data control No server, sync via your private cloud, direct connections.
Request beta