honcho

To maintain AI agent memory across multiple sessions, configure per-profile memory management using Honcho. This models users per profile, isolates AI peers, and controls observation and recall settings to inject session summaries into prompts.

Set up cross-session recall by running the setup wizard to configure a per-profile honcho.json file. You then verify memory status using standard CLI tools to ensure the backend supports your workspace's hybrid, context, or tools recall modes.

Cross-session user modeling is the process of maintaining a coherent user representation across conversations. You need it when your AI agent must retain context and dialectic reasoning depth without exceeding context budgets.

Yes, you can control what an AI agent observes and recalls using configurable observation controls and recall modes. You can apply per-directory or per-session scope to tailor exactly what is learned and recalled.

Yes, managing per-profile memory requires a working Honcho backend and a valid SKILL.md frontmatter with a name and description. You also need standard Rawclaw CLI tools to execute setup, status checks, and synchronization.

Context budget limitations restrict the volume of session summaries and user representations injected into prompts. You manage these constraints by adjusting dialectic reasoning depth controls and selecting appropriate recall modes for scalable performance.