agent-middleware

Open-source middleware for AI agents — catches mid-run failures,compresses stale context, and grows a reasoning library across runs. Any model, any framework.

Layered token-optimization pipeline for DeepSeek Harness: output ladder, MCP lazy loading,compaction driver, cache-hit reporting. Built on real DSH plugin APIs; ~40-60% input saved in long sessions.

Behavior-management plugin for DeepSeek Harness: tool-call discipline prompt section, failure-triggered parallelism convergence (pool drops to 1, auto-restores), consecutive-failure user intervention. Complements dsh-token-optimizer; works standalone.

Open-source agent middleware — shared memory, credential isolation, and schema learning for AI agents. What one agent learns, every agent knows.

Mentalist: (DeepAgent in rust) High-performance Rust agentic middleware & execution environment for autonomous AI agents. Implements DeepAgent methodology with execution hooks, explicit planning, sandboxed tool execution, and full session serialization. Built for production-ready agent systems.

TikTok TechJam 2026 Track 1 — Glass Box: agent-observability middleware turning every Agent Run into a correlated span waterfall

Framework-agnostic confidence-gated escalation middleware for LLM agents: multi-signal scoring (logprob, verbalized, tool risk), threshold policies, and escalation handlers for LangChain, CrewAI, AutoGen, and Google ADK.

LangChain agent middleware for Cycles — pre-tool-call authorization and per-tenant budget enforcement for Python agents using create_agent