Python

An open-source reinforcement learning training loop for LLMs, spanning small CPU-friendly experiments like CartPole to async RLVR runs on GPU clusters. The repo highlights a compact codebase, quickstart commands, task customization, and support for algorithms like REINFORCE, PPO, GRPO, and RLOO.

A research repo for reducing hallucinations in LLM-generated code using semantic triangulation, with setup, benchmarking, experimentation, and reproducibility instructions.

Hubmesh is a Python library for multi-hop GraphRAG retrieval planning on top of existing vector databases. It adds centrality-aware seed selection, budget-aware context packing, entity graph building, and an MCP server for agent/tool integrations.

Open-source Python library for AI agent context governance with tamper-evident audit trails, provenance validation, freshness checks, and adapter support for common agent stacks.

A Python package and CLI for checking how closely an AI-generated answer reuses its source text, with support for verbatim matching, subsequence matching, quotation verification, and pipeline gating.

A local terminal UI for exploring AI coding spend across tools like OpenCode, Claude Code, Codex, GitHub Copilot, and more. It shows costs by month, day, project, session, and model, with drill-down views, heatmaps, exports, and a web browser mode.

A deterministic Python project for empirical law discovery in elementary cellular automata, with simulation, discovery, reproducibility guides, and published preprints.

A Python library that transpiles stateful imperative LLM workflows into optimizable DSPy programs, with typed outputs, self-correction loops, validation reporting, and prompt tuning support.

Recall is a local-first memory tool for Claude Code that logs sessions and generates a compact project summary offline, without sending data to an API.

MakerChecker is a self-hosted governance layer for AI agents that enforces roles, approvals, and tamper-evident audit logs. The repo shows quickstart commands, a demo flow, and SDK/connectors for TypeScript, Python, and LangChain-style integrations.

An MCP server for SharkClean and SharkNinja robot vacuums that lets MCP clients start cleanings, target rooms, dock the robot, and check status. The repo also includes setup steps, authentication guidance, CLI commands, and notes on supported models and cloud backend behavior.

Nanocode is a terminal-based AI coding assistant written in Python. The repo shows CLI install and usage instructions, runtime commands, configuration options, supported providers, and a detailed description of its context and file-editing workflow.

A pure-Python LLM inference engine and server with CUDA/HIP/Metal code generation, OpenAI-compatible API support, built-in RAG, and multi-GPU backend support.

tokentoll is a GitHub Action and CLI that scans Python and JavaScript/TypeScript code for LLM API calls, estimates spend, and blocks pull requests when cost or model-policy thresholds are exceeded.

SLayer is an open-source semantic layer for AI agents and humans that sits between a database and downstream tools. It can auto-create data models from schema, compile natural-language-style queries into SQL, and be used through MCP, REST, CLI, or Python.

Open-source Python framework for building multi-agent LLM apps that can run on local models through Ollama or on hosted APIs like OpenAI, Anthropic, and OpenRouter. The repo highlights built-in plugins, workflow orchestration, agent serving over HTTP, and guardrails for tool use.

A Claude Code status line that shows session size, next-turn cost, and remaining budget before you type.

FastAPI-based server that turns files and Python modules into REST APIs, browsable UIs, searchable resources, and an MCP endpoint.

Local-first debugger for LLM agent context windows. It records each call’s context into a single-file trace so you can inspect, diff, and share what the model actually saw turn by turn.

Local CLI for analyzing when a coding agent starts degrading as session context fills, using transcript data already stored on disk.

AutomatiQ is a Python CLI that records a browsing session and uses vision and LLM steps to generate a standalone HTTP-based automation script.

An open-source Python support agent that uses SynapCores as the agent brain for memory, retrieval, semantic tool routing, and generation. The repo includes a quick start, local Docker setup, example commands, and a traced demo of the agent loop.

Klent is an AI agent guardrails product that sits between agents and their tools, letting teams allow, deny, pause, rewrite, or redirect tool calls with policies, human approval flows, and audit trails.

Open-source harness for executable security regression testing of agentic applications and MCP-integrated systems, with scenario files, policy assertions, trace-based evaluation, and JSON output for CI or local use.