TensAI Reusables
59 curated open-source repos, each solving one sub-problem from the Spine. Clone the one that matches your need. Every repo maps to a concept page that teaches the context.
Universal memory layer — semantic, episodic, factual
Temporal knowledge graph memory
Persistent memory for coding agents via MCP
Lifelong memory with semantic compression
Self-organizing dynamic memory
Self-improving context via reflection
Interaction control harness for customer-facing agents
Full Agent = Model + Harness reference
Harness engineering tutorial
Lightweight graph RAG
Self-correcting / corrective RAG
Multi-strategy chunking library
Context-preserving chunk embeddings
Unified reranker API
RAG evaluation framework
Codebase knowledge graph (tree-sitter + MCP)
Incremental data indexing engine
Semantic code search (ripgrep + tree-sitter)
High-perf code intelligence (155 languages)
Official MCP reference server implementations
High-level TypeScript MCP server framework
Production MCP server for databases
Containerised MCP server management
Agent-to-MCP integration framework
Anthropic's canonical five pattern implementations
Five patterns + MCP + Temporal durability
100-line minimal agent/workflow core
Kill-switch, policy enforcement, OWASP coverage
Self-hosted background agent runner
Teams-first multi-agent orchestration
Agent credential management with kill-switch
Prompt/agent/RAG testing with CI/CD
OpenTelemetry-native LLM observability
Agent trajectory evaluators
Snapshot-based agent regression testing
Control vectors via representation engineering
Structured outputs (Python constrained decoding)
Fast grammar-based constrained decoding
RepE for hallucination/safety control
Agent failure diagnosis from trajectories
Multi-agent failure taxonomy
Real-world agent failure case studies
LLM vulnerability scanner (red-teaming)
LLM input/output security scanner
Static security scanner for agents (OWASP)
Kubernetes-native agent sandbox