honcho

Cross-session user context is maintained by building a persistent user model that represents who the user is and applying it consistently to each agent profile. This prevents losing conversational continuity across turns.

Configuring Honcho memory requires installing the honcho-ai package, running honcho setup to input your API key, and defining memory provider settings in your honcho.json host configuration. This enables consistent user representation.

Multi-profile isolation allows each agent profile to maintain its own AI peer identity while sharing a unified user context across all profiles. This ensures tailored responses without cross-profile interference.

Available recall modes include hybrid, context, and tools modes. These control how memory is applied during context injection, letting you tune what is learned and how synthesized memory is used.

Context budget enforcement prevents prompt bloat by trimming lower-priority memory components, such as summary or representation data, when exceeding configured token budgets. This keeps context injection efficient.

Dialectic reasoning controls allow tuning cadence, depth, and reasoning level to balance answer quality against cost. These settings manage how memory-based answers are synthesized during recall.