wipnote

ai-engineering
multi-agent
infrastructure
developer-tools
Published

February 1, 2026

Overview

Status: Active development (Go binary) Impact: Dogfoods its own development — all wipnote work items tracked in wipnote Technologies: Go, HTML, SQLite, Claude Code Hooks Docs: shakestzd.github.io/wipnote

Local-first observability and coordination platform for AI-assisted development. wipnote (originally launched as HtmlGraph) gives AI coding agents (Claude Code, Gemini CLI, Codex, GitHub Copilot) a shared coordination layer. Work items are canonical HTML files in .wipnote/, queryable via SQLite, managed by a single Go binary. No external infrastructure — just your repo.

Architecture

Three layers:

  1. HTML file storage — Work items stored as .wipnote/*.html files in your repository. HTML is the source of truth: human-readable, git-diffable, browser-renderable.
  2. SQLite cache — Enables efficient querying, dashboard functionality, and analytics without parsing HTML on every request.
  3. Go binary — Single CLI binary for work item management, hook execution, and local dashboard serving.

Key Features

Work item management — Tracks features, bugs, spikes, and initiatives using prefixed identifiers (feat-, bug-, spk-, trk-). Items progress through defined lifecycle stages: creation, active work, and completion.

Session tracking — Tool calls and prompts are logged and connected to specific work items, creating an audit trail viewable through the dashboard.

Agent customization — Configure specialized agents with distinct models, tools, and instructions for different roles (research, coding, testing).

Automation hooks — Event-driven triggers respond to session start, tool execution, and completion events to enforce safety rules or initiate workflows.

Reusable commands — Slash command shortcuts encapsulate multi-step processes into single invocations.

Quality enforcement — Built-in gates require successful builds, linting, and testing before commits.

Dashboard — Locally-served web interface with real-time activity monitoring, kanban views, and work item details.

Multi-tool coordination — Claude Code, Gemini CLI, and other AI coding tools can read and write the same work items.

Design Philosophy

  • HTML files as structured graph storage — browsers render them, git versions them, SQLite queries them
  • Zero external dependencies — no databases, no cloud services, no accounts
  • Single binary distribution — install and run
  • AI agents are first-class citizens, not afterthoughts
  • The filesystem is the database, the browser is the UI

Documentation & Resources