agent-wiki-template

agent-wiki-template.vercel.app — landing page with a full overview for all audiences.

A structured knowledge base built for AI agents, not humans. Clone this repo, connect your data sources, and run one bootstrap prompt to generate a living wiki from your operational history.

What this is

Most teams run on Slack, meetings, and scattered docs. Knowledge lives in people's heads and disappears when threads scroll past. This template gives you a framework to compile that knowledge into structured articles that an AI agent can read, reference, and update.

The wiki is designed to be:

• Read by agents as context before they do work (briefings, analysis, planning)
• Updated by agents as new decisions are made
• Auditable by humans when they need to check what was decided and when

Token efficiency

The core pattern is state.md as a compressed snapshot layer.

Without state.md: agents fetch index.md + matching workstream files + people files at Step 0. On a mature wiki with 10+ workstreams, this costs 30-40K tokens per run before any live data is pulled.

With state.md: agents fetch one ~800-token file. Individual files are only fetched when a specific gap forces it. On the same wiki, Step 0 costs 800-1,500 tokens.

state.md is regenerated at the end of every briefing run. It is always current as of the last evening cycle. It is never manually edited.