Agent Memory Wiki

A research agent reads its MEMORY.md at session start, sees that Tavily's pricing was last verified on 2026-05-01, and skips re-checking it. When it discovers HasData's new Business tier ($99/1M), it appends the finding to the wiki with a date stamp. Next session, that knowledge is immediately available.

Agent Memory Wiki is relevant to Google. Scavio provides a unified API to access data from all of these platforms.

AI agents lose all context when their conversation ends or their context window fills up. Agent memory wikis solve this by giving the agent a structured Markdown file (or set of files) that it reads at the start of each session and updates as it learns new information. The wiki format works well because Markdown is human-readable, git-diffable, and token-efficient compared to database schemas. The practical pattern: the agent reads a MEMORY.md file at session start, uses it to inform decisions, and appends or edits entries when it discovers new facts. Because the file lives in a git repo, every change is versioned, reviewable, and reversible. This is the approach used by Claude Code's CLAUDE.md and similar systems. The key design decisions are: what to remember (facts vs preferences vs workflow state), how to structure entries (flat list vs categorized sections), and when to prune (manual vs automated summarization). For search-heavy workflows, the agent memory wiki stores things like verified pricing for APIs, known-good query patterns, provider reliability observations, and domain-specific knowledge that would otherwise be lost between sessions. Teams using Scavio for research workflows store verified data points (pricing tiers, API endpoints, feature availability) in the wiki so the agent does not need to re-verify them in future sessions.