Playbooks: Agent Integration

This document provides step-by-step playbooks for integrating AgenticCognition into agent workflows. Each playbook includes setup instructions and usage patterns.

Playbook 1: Personal Assistant Agent

Connect cognition to a personal assistant to provide responses calibrated to the user's belief system, decision patterns, and growth trajectory.

Setup:

1. Install AgenticCognition and configure the MCP server for your client
2. Create a living user model: acog model create
3. Add initial beliefs based on onboarding conversation

Usage pattern:

• On each interaction, call cognition_model_heartbeat with the conversation topic
• Before giving advice, call cognition_predict to check alignment with user values
• Periodically call cognition_model_portrait to refresh the assistant's understanding
• When the user faces a choice, call cognition_simulate to evaluate options

Playbook 2: Coaching Agent

Use shadow psychology and drift tracking to build a coaching agent that surfaces blindspots, tracks growth over time, and provides personalized guidance based on self-concept topology.

Setup:

1. Create a dedicated coaching model with acog model create --name coaching
2. Add beliefs across multiple domains: values, relationships, work, growth
3. Build initial belief graph with entanglement links

Usage pattern:

• Call cognition_shadow_map monthly to check for new projections or blindspots
• Use cognition_drift_track to show the user how their values have shifted
• Call cognition_self_topology to identify defended regions that may limit growth
• Present shadow insights tentatively, giving the user agency over engagement
• Track growth rings to celebrate progress over time

Playbook 3: Decision Support Agent

Combine decision simulation with belief graph analysis to help users make better decisions.

Setup:

1. Ensure the model has beliefs in the relevant decision domain
2. Build entanglement links between related beliefs (support, tension, opposition)
3. Establish keystone beliefs that anchor the domain

Usage pattern:

• When the user presents a decision, call cognition_simulate with the scenario and options
• Display per-option alignment scores showing how each choice relates to core values
• Call cognition_belief_graph to visualize which beliefs support or oppose each option
• Use cognition_pattern_fingerprint to show the user their decision-making tendencies
• After the decision is made, record the outcome to refine future predictions

Playbook 4: Multi-Agent Ecosystem

In a multi-agent setup, AgenticCognition serves as the shared understanding layer.

Setup:

1. Create a single model that all agents share via the same .acog file
2. Configure each agent's MCP server to point to the same storage directory
3. Use file locking to prevent concurrent write conflicts

Usage pattern:

• Each agent calls cognition_model_heartbeat to record its interactions
• Agents call cognition_predict before making decisions on behalf of the user
• The coordination agent calls cognition_model_vitals to monitor model health
• All agents benefit from the accumulated understanding of every other agent
• Use cognition_soul_reflect as a shared context briefing for new agents

Playbook 5: Team Context Agent

Use per-project .acog files to maintain separate user models for different work contexts.

Setup:

1. Set ACOG_STORAGE to a project-local directory: export ACOG_STORAGE=./.acog
2. Create a project-specific model: acog model create --name "project-x"
3. Add beliefs relevant to the project context

Usage pattern:

• The agent automatically loads the correct model based on the active project directory
• Work-domain beliefs are kept separate from personal beliefs
• When switching projects, the agent's understanding switches with it
• Use cognition_drift_track to monitor how the user's relationship with a project evolves
• Archive completed project models for future reference