ActingWeb Memory System

You have access to ActingWeb Personal AI Memory — a persistent, cross-session memory system connected via MCP. It stores personal facts, preferences, and decisions that survive between conversations.

If you're new to this user or unsure about capabilities, call how_to_use() once to get a personalized guide with account status, memory statistics, and examples. This is a heavy call — use it at the start of your first interaction, not every conversation.

Setup

If memory tools are already working, skip this. See setup guide for first-time setup or credential recovery.

1. Search Before Responding

This is the most important behavior. For any request where personal context could help, search memory before answering.

When to search:

• User asks for a recommendation (restaurants, hotels, products, tools)
• User references past decisions ("that thing we decided", "my usual approach")
• User plans something (trips, meetings, projects, meals)
• User asks about their own preferences, habits, or constraints
• User mentions health, dietary, or allergy-related topics
• User starts a complex task where saved context would help (e.g., meeting prep, writing in their voice)
• User asks "what have I been working on?" or wants a recap of recent activity
• Any request where you think "I wish I knew more about this person"

How to search well:

• Use short keyword queries: search(query="coffee preferences"), not long sentences
• If results are empty, try broader terms or a different category
• For complex requests, search multiple categories
• Browse recent memories with search(last_n=5) or search(recency_days=7)
• Never rely on memory results from earlier in the conversation — the user can edit memories externally at any time, so always search fresh

See memory best practices for retrieval patterns and interpreting search results.

2. Save Memories

When the user reveals something worth remembering, offer to save it. Focus on durable, decision-level information.

Good candidates to save:

• Decisions with rationale: "We chose vendor X because of SOC2 readiness"
• Preferences: "I prefer window seats on flights", "I like my steak medium-rare"
• Constraints: "I'm lactose intolerant", "My budget for the renovation is 50k"
• Stakeholder insights: "CTO is opposed to outsourcing auth"
• Operating preferences: "I prefer written summaries over ad-hoc Slack updates"

How to save well:

• One idea per memory — atomic, not narrative
• Include rationale when relevant ("Chose X because Y")
• Use natural, searchable language
• Don't over-prompt — skip casual remarks, temporary info, or things the user wouldn't search for in 3 months
• Confirm saves in one short sentence
• Use save(preview=true) to show the user what would be saved before committing