# Fabric blog

> Engineering notes from the team building Fabric.

Canonical: https://fabric.pro/blog

---

- [Building a Dynamic Agent Loop with Temporal: Durable Orchestration for Tools, Workflows, and Workspaces](https://fabric.pro/blog/building-dynamic-agent-loop-temporal) (2026-03-01): How Fabric's orchestrator became a Temporal agent loop with on-demand tool discovery, MCP, workspace retrieval and human approval steps.
- [How AI Agents Learn and Remember: Building a COALA-Inspired Memory System](https://fabric.pro/blog/agent-memory-system-how-ai-agents-learn-and-remember) (2026-01-16): How Fabric gives its agents and orchestrator memory across conversations. It uses four tiers, semantic search, automatic episodes and context injection.
- [Hybrid tool search in Fabric's orchestrator](https://fabric.pro/blog/intelligent-tool-search-orchestrator) (2026-01-15): When an agent can reach hundreds of tools, sending them all to the model wastes tokens. This is how Fabric picks the right few with keyword and meaning search.
- [Building a Production-Grade Database-Driven AI Model Selection System](https://fabric.pro/blog/ai-model-selection-gateways-providers-routing) (2026-01-12): How Fabric moved every AI model choice into the database, with tenant isolation, provider routing per task and no model names hardcoded anywhere.
- [Deep Dive: How Fabric AI's RAG Architecture Reads Enterprise Knowledge](https://fabric.pro/blog/deep-dive-rag-architecture-document-intelligence) (2026-01-05): How Fabric extracts documents, chunks them by meaning, embeds them in Qdrant and retrieves with reranking, so agents answer from a team's own data.
- [Introducing Fabric AI: The Enterprise Platform for SDLC Compression](https://fabric.pro/blog/introducing-fabric-ai-sdlc-compression-platform) (2026-01-04): Why Fabric exists: documents, tickets and decisions scattered across tools slow every release. How retrieval, agents and durable workflows shorten it.
- [Understanding Fabric Agent Types: Orchestrator, Composable, and Custom Agents](https://fabric.pro/blog/understanding-fabric-agent-types) (2026-01-03): A guide to the three kinds of AI agent in Fabric, when to use each, how they work, and practice for building systems of several agents.
- [Building the Fabric Orchestrator with Temporal](https://fabric.pro/blog/fabric-orchestrator-temporal-multi-agent-orchestration) (2026-01-02): Fabric's orchestrator runs on Temporal. It routes work between agents, survives restarts, remembers past runs and asks people at the right moment.