Vibeleaderboard
Index / agent
Visit github.com
Category
AI Agents
Rank
No. 1713Tools index

Previous survey · No. 1691 ·

Pricing
Open Source
Type
AGENT
Use case
Coding · Workflow Automation
Interfaces
CLI
GitHub
8 stars
Latest release
v0.14.0
Date

About

Kana is an open-source, single-binary terminal AI agent for coding, system workflows, and tool-driven productivity. It offers an interactive TUI that keeps reasoning, diffs, and execution together, a headless "kana exec" mode for scripts and CI pipelines, durable memory across sessions, and reusable templates and subagents, all without telemetry.

At a glance

Kana is a terminal coding agent that can retain project conventions between sessions. It is worth a scratch trial if you want to inspect tool calls and see which project rules carry over after a restart.

Good fit

macOS and Linux developers who want a terminal agent with configurable model endpoints and persistent context.

Consider something else when

Skip it if shell commands and file access need to be isolated from your host.

Start here

In a disposable workspace, give Kana a harmless project convention and inspect it with /memory. Start a fresh session and check whether it recalls the convention before relying on it for real work.

Cost and requirements

MIT-licensed. The project lists prebuilt binaries for macOS and Linux on arm64 and x64. You need a configured model endpoint and any credentials it requires. Hosted providers may charge for calls.

Limits to know

Session history and memory stay on disk and may contain sensitive material. The model endpoint receives request context. Approvals are per-tool decisions and commands still run with your host permissions.

Why it made the leaderboard

Offers a terminal coding-agent alternative built around persistent memory and no telemetry for developers who want a Claude Code/Codex-style workflow without usage data leaving their machine.

Tags

ai-agentsterminalcodingcliopen-source

Tech Stack

Node.jsTypeScript

Comments (0)

No comments yet

Editorially curated, with community endorsements as a secondary signal. Corrections welcome.