About Kaizen Manas
Kaizen Manas is a personal AI agent harness that you control from Telegram or Discord. The agent inside it is called Manas. It runs on a small Python host and asks you before it does anything risky.
What it is
A harness is the code around a language model. It receives chat messages, gives the model a set of tools, checks risky actions and keeps memory.
What it does
- Chats through Telegram or Discord.
- Offers 61 tools in 13 groups. Five are always on.
- Runs Python in a limited sandbox. Risky code waits for a button.
- Keeps notes, a small graph, vector search and an event log.
- Runs scheduled prompts.
- Routes work between a small, a medium and a power model.
What it does not do
- It does not host a model. You bring your own key.
- It is not a container platform. The sandbox is a limited subprocess.
- It does not read photos or files you send.
- No performance numbers exist.
Who made it
Adarsh Jaiswal. The code is MIT licensed. The source repository is private.
In this section
PositioningWhere Kaizen Manas fits and what it is not.AudienceWho this documentation is for.Name and historyOfficial spelling, names, rename date and key facts about the repository.Design principlesThe ten rules the design follows, and why.
See also:PositioningName and history
Checked against the source at commit f4fd89f on 29 Oct 2026. Everything here is merged on the main branch; no release has been tagged yet. The source is private, so these pages are the public description.