Local-first · Read-only · Zero telemetry

Your local AI agent assets, finally visible.

Berth scans the Skills, MCP servers, hooks, subagents and sessions scattered across your machine and connects them into one browsable, searchable map — so you finally know what you have, how it works together, and what it costs.

Download BerthView on GitHub

macOS 11+ / Windows 10+ · Free & open source (MIT)

berth · ~/.claude
Capabilities25 assets
  • code-reviewSkill
  • githubMCP
  • PreToolUseHook
  • reviewerSubagent
  • commit-helperSkill
Read-only — never changes your filesRuns fully on your machineNo telemetry, no accountCredentials never shown
Why Berth

Invisible assets, unmanaged capabilities

Make assets visible

Turn plain-text skills, hooks and MCP servers from a scattered file tree into structured, connected objects.

One view across agents

Configuration, capabilities and sessions for Claude Code and Codex, side by side in one place.

Local-first

Everything runs on your machine — zero telemetry, no cloud sync, no account required.

Read-only & safe

v0.1 never modifies a file. API keys and tokens are detected, never displayed.

Core features

Your whole agent setup, at a glance

Overview

A dashboard for everything

Asset counts, recent sessions, this week's spend and health checks — all on one screen.

  • Asset inventory
  • Recent activity
  • Cost & health checks
01
Asset inventory
02
Recent activity
03
Cost & health checks
Sessions

Browse past work

Walk through sessions by project or date, and see the skills loaded, MCP servers connected, hooks fired and artifacts produced.

  • Group by project or date
  • Loaded assets per session
  • Tool timeline
01
Group by project or date
02
Loaded assets per session
03
Tool timeline
Configuration

Take inventory

Instructions and Capabilities in two columns, merged across user, project and enterprise scope, with sources and conflicts marked.

  • Instructions & Capabilities
  • Scope merge
  • Conflict detection
01
Instructions & Capabilities
02
Scope merge
03
Conflict detection
Usage

Spend & tokens

Cost and token trends by model, project and day, with rate-limit headroom kept in plain sight.

  • By model / project / day
  • Token breakdown
  • Rate limits
01
By model / project / day
02
Token breakdown
03
Rate limits
Knowledge base

First, understand what AI agents can do

Before managing these assets, get a plain-language grasp of what makes an agent an agent — perception, planning, tool use, memory, and autonomous multi-step execution. Every explanation cites a primary source.

Enter the knowledge base
Three tracks

Not a feature dump — a way in to understanding agents

Learn

Understand AI agents

In plain words and trusted sources: how an agent differs from a chat model, and its core capabilities.

Start learningFeatures

Berth features in depth

A walk through Overview, Sessions, Configuration and Usage — and the asset model behind them.

Explore featuresGuides

Hands-on guides

Diagnose why a hook isn't firing, make sense of your cost, set a config baseline for your team.

Follow along
FAQ

You might be wondering

Will Berth modify my files?+

No. v0.1 is fully read-only — it scans and displays, and never writes to any config or asset file.

Is my data uploaded anywhere?+

No. Berth runs entirely on your machine: zero telemetry, no cloud sync, no accounts.

Which agents are supported?+

v0.1 focuses on Claude Code, with an adapter layer already designed for Codex and more in later releases.

Is it free?+

Yes. Berth is open source under the MIT license — contributions welcome on GitHub.

See your AI agents clearly — start with Berth

Free, open source, local-first. Download and go, or learn more on GitHub.

Download BerthStar on GitHub

macOS 11+ / Windows 10+ · MIT