SourcesPlans and work
Mithrandir + AI agent chats
Connecting AI agent chats keeps the reasoning people work out with their agents instead of losing it when the session ends. Claude Code sessions and other agent histories become part of the company memory. Every conversation stays private to the person who had it, unless they share it with a named colleague.
What it reads
What Mithrandir reads from AI agent chats
- Claude Code sessions
- Chat history from any agent that saves it in a standard format
- Who said each turn, and which model answered
Access
The access rules that carry over
- Every conversation is private to the person who had it, by default.
- The owner can share one conversation with a named colleague, and take it back.
- Keys and passwords pasted into a chat are masked before anything is kept.
Setup
Set up in three steps
- 1
Each person points Mithrandir at where their agent keeps its history.
- 2
Their conversations come in, private to them.
- 3
They ask from any agent they use and get answers from their own past work, and from conversations shared with them.
Examples
What your agents can now answer
- “How did I fix the deploy error last week?”
- “What did I decide about the data model in yesterday's session?”
- “What approach did my teammate take in the conversation they shared with me?”
Questions
AI agent chats and Mithrandir, answered
- Can my manager read my agent chats?
- No. Your conversations are private to you. A colleague sees one only if you share it with them.
- Which agents are supported?
- Claude Code, and any agent whose history can be saved in a standard format.
- What about secrets pasted into a chat?
- Keys and passwords are masked before a conversation is kept, so they never reach an answer.
Last updated 25 October 2026 · Mert Uzun, co-founder
See it on your own AI agent chats.
Thirty minutes with the founders, one workflow, no form.
Book a call