Cognition Engines Introduces Decision Intelligence for AI Agents

Cognition Engines introduces a decision intelligence platform specifically designed for AI agents, aiming to enhance their autonomy with structured reasoning, historical context, and enforced guardrails. The core functionality revolves around enabling agents to "Query Before Deciding," leveraging semantic search across past decisions to identify successful patterns, failures, and underlying rationales.
The system supports directional search, allowing agents to query by structure (e.g., "where did we use this pattern?") or function (e.g., "what solved this problem?"). This capability is crucial for building a compounding record of organizational judgment, ensuring consistency and learning across agent operations.
The operational framework is built around a five-phase loop: Fetch (load context and past decisions), Orient (check guardrails and constraints), Resolve (decide and record with reasoning), Go (execute), and Extract (evaluate outcomes and distill patterns). This systematic approach ensures that every decision is informed, constrained, and documented.
Cognition Engines is available as a Claude Code plugin, offering hooks, commands, and skills to automate the entire decision loop. The workflow is streamlined into two primary API calls: `get_session_context` at session start to retrieve cognitive context (profile, calibration, patterns), and `pre_action` at decision points to handle query, guardrails, and recording in a single call.
The platform automatically captures deliberation traces, bridge-definitions, and related decision links without requiring client-side changes. The output includes critical information such as `allowed` status, `similar_decisions`, `guardrail_results`, and `calibration_context` (e.g., Brier score, accuracy), along with a unique `decision_id`.
Version v0.14.0 introduces significant enhancements, including Multi-Agent Isolation for managing distinct agent environments, Live Deliberation for real-time decision analysis, and a FORGE Plugin for extended capabilities. These features collectively aim to provide a robust, auditable, and intelligent decision-making layer for complex AI agent deployments.
The strategic implication of Cognition Engines is its potential to elevate the trustworthiness and explainability of AI agents. By providing a transparent and verifiable decision-making process, it addresses key concerns around AI governance, safety, and compliance, making advanced AI agent systems more viable for critical applications.
EU AI Act Art. 50 Compliant.