Connecting docs

In ctx|, documentation is primarily repository-native: the same Git connection that powers search and ingestion also surfaces the files agents and graphs can reason about.

What counts as “docs”

• Top-level guides — README.md, CONTRIBUTING.md, docs/**
• Architecture decisions — ADR-*.md, .ai/memory/decisions/** (if present)
• Agent instructions — AGENTS.md, .cursor / .agents skills (depending on how your org layouts them)

When ctx_advisor runs, the stack is designed to treat that material as part of your instruction hierarchy, not as an afterthought pasted into chat.

Practices that work well

• Keep one source of truth in Git; avoid duplicating policy in Slack-only posts if you want agents to cite it.
• Prefer small, linked ADRs over giant wiki pages — easier to ingest and attach to graph claims.
• Use AGENTS.md at repo (or monorepo) roots the way you would for human contributors: concrete, scoped, maintained.

Hosted vs self-hosted

The OSS deployment does not ship a separate “Google Docs connector” in these docs. If you need external doc systems, typical patterns are: export/sync into the repo, or integrate via your own automation pushing into ctx| APIs — treat that as custom glue until a first-class connector exists in product.