llm-agents

MCP server that lets AI agents control labgrid-managed embedded hardware labs through 47 tools, including power, console, flashing, muxing, SSH, and place management.

Runtime safety net for LLM agents that detects runaway token growth, classifies failure patterns, and suggests fixes without extra LLM calls. The project includes a Rust core, Python SDK, examples, benchmarks, and benchmark claims across multiple models and runs.

Leviath is a Rust-based agent runtime for running structured LLM agents from a single file. It emphasizes context regions, stage-specific models and tools, run journaling, live steering, and the ability to run many agents in one process.

A local sandbox for running LLM-generated code from Python or Node with no cloud, no daemon, and strong isolation controls.

A TypeScript library for deterministic context folding in long-running LLM agent sessions, designed to preserve prompts, avoid extra summarization calls, and keep provider caches hot.

Go-native framework for building, orchestrating, and observing AI agents with native OpenTelemetry, an embedded dashboard, replay, and first-party MCP/A2A support.

Go-based multi-source search CLI for LLM agents with session-based workflows, paginated result reading, and support for web, HN, Reddit, GitHub, YouTube, and multiple search APIs.

LLM Wiki is a plugin and instruction-file based system that turns LLM agents into a research and knowledge-base workflow with session memory, parallel research, source ingestion, and wiki compilation.

OpenAPPA is an open-source deterministic guardrail and information-flow policy engine for LLM agents. The page describes flow-based security labeling, CI validation, remedy plans, sanitizers, authorities, and subagents, with links to docs, GitHub, Discord, and benchmark claims.

KHMS is a file-based long-term memory system for LLM agents, built around immutable markdown cards in a git repository with YAML frontmatter, hook-based recall, and a propose → review → approve flow.

A self-hosted TypeScript harness for building provider-neutral LLM agent systems inside an application or platform, with typed agent and workflow runtime, tool support, state, tracing, and example workspaces.

A no_std Rust memory engine for on-device LLM agents that uses provenance-aware recall to keep a model’s own output from becoming evidence.

A service-boundary-aware coordination layer for heterogeneous LLM code agents via MCP. It combines versioned Markdown documents, pub-sub notifications, diff-aware updates, and a web dashboard for browsing, search, and AI chat.

A deterministic security scanner and adversarial flight simulator for AI agents, built for red-teaming LLM workflows and MCP-compatible setups.

A type-safe Go runtime for building tool-using LLM agents across multiple providers, with streaming, memory, and OpenTelemetry support.

A source-level reconstruction of the WON-era Half-Life launcher, build 1792, with build instructions, supported-version notes, fixes, and a bundled WON server emulator for local testing.

A technical guide that shows how to build a small CLI-based LLM agent for microservices using tool discovery, model calls, conversation history, and a REPL loop.

A reference implementation for intent-execution separation in LLM agents. The project uses a typed intent layer (SIF) over domain vocabularies, with a deterministic execution layer and demo domains for legal and veterinary workflows.

A research article describing an empirical runtime-monitoring approach for multi-turn LLM agents, backed by 3,175 runs across four benchmarks. It also points to an open-source implementation, state-harness, with Rust/Python support, LangGraph and CrewAI adapters, CLI tooling, and OpenTelemetry export.

arXiv paper on self-speculating LLM agents that predict their next tool call to hide tool latency and improve next-call Hit@1 while preserving task success.

An arXiv paper on reducing LLM inference cost for web automation by compiling browser tasks into a deterministic JSON workflow and executing them without repeated model calls.

arXiv paper proposing MANTA, a framework that lets multi-agent communication structures adapt at inference time based on collaboration traces and task needs.