3/2/2026
Show HN: Llmdoc – annotate codebase with LLM summaries only re-scan what changed
CLI tool that adds LLM summaries to codebase files and intelligently re-scans only changed files to reduce token costs.
3/2/2026
Show HN: OnCallMate – AI agent for autonomous Docker incident RCA
OnCallMate: autonomous AI agent for Docker incident detection and RCA, monitoring containers, detecting anomalies, and suggesting fixes via function calling.
3/2/2026
Show HN: Watchtower – see every API call Claude Code and Codex CLI make
Real-time monitoring dashboard for API calls between AI coding agents (Claude Code, Codex CLI) and their APIs. Network inspection tool for agent debugging.
3/2/2026
Show HN: Accel-GPU – NumPy for the browser GPU
WebGPU library providing NumPy-like API for GPU arrays in browser with fallback to WebGL2/CPU. 160KB, zero dependencies, includes FFT and attention operations.
3/2/2026
Show HN: Ccmux – Reduce context switching for parallel Claude Code sessions
Tmux wrapper (ccmux) for managing multiple parallel Claude Code sessions with unified UI sidebar and shared git worktree management.
3/2/2026
Show HN: VS Code AI copilot that catches mistakes before they break your code
VS Code AI copilot enforcing structured specs with review-first workflow requiring human approval before code patches are applied.
3/2/2026
I built a persistent memory layer for AI agents in Rust
Persistent memory system for AI agents in Rust with Python CLI. SQLite backend storing text, vector embeddings, and metadata with hybrid search. No cloud dependency.
3/2/2026
Show HN: Vim-Claude-code – Claude CLI integration for AI workflows inside Vim
Vim/Neovim plugin integrating Claude CLI for AI-assisted development workflows including code refactoring, test generation, and commit message improvement.
3/2/2026
Ask HN: Git branching strategy when using multiple CLIs running multiple agents?
Discussion question about git branching strategies when managing multiple concurrent AI agent instances across terminal sessions.
3/2/2026
Show HN: Smidge. Turn expert knowledge into agent intelligence
Smidge converts source materials into production-ready agent skills compatible with Claude, Cursor, and 30+ AI agents using source-aware extraction.
3/2/2026
Run your agent 10 times – you won't get the same answer
Research on LLM agent output consistency: running same agent multiple times produces different results despite identical inputs.
3/2/2026
Show HN: Nanograph – On-Device GraphDB for Agents
Nanograph is an on-device graph database designed for AI agents.
3/2/2026
Show HN: Git-stint – Built for AI coding agents to multitask without collisions
Git-stint automates workflow management for multiple parallel AI coding agents on shared repos, preventing git collisions with isolated sessions.
3/2/2026
Show HN: Built lovable but for your existing products
AI-powered feedback widget for Next.js that converts user chat into GitHub issues and autonomous agent implementations with self-learning.
3/2/2026
Compiling English Security Policies into Deterministic Agent Guardrails
Research on compiling plain English security policies into deterministic guardrails for autonomous AI agents without runtime LLM involvement.
3/2/2026
Show HN: Memgraph-agent – NER+PageRank memory for AI agents, $0 LLM cost
Graph-based memory system for AI agents using NER and PageRank for zero LLM cost context management. Open-source Memgraph integration.
3/2/2026
On-Policy Prediction with Approximation
Machine learning research on on-policy prediction with function approximation using balance and visitation equations.
3/2/2026
Show HN: 58% cost by replacing file reads with a dependency graph on AI Coding
MCP server using tree-sitter dependency graphs to reduce token costs in AI coding agents by 58%. Includes benchmark results on Claude with 42 test executions.
3/2/2026
A case for Go as the best language for AI agents
Article arguing Go's concurrency model, performance, and deployment simplicity make it ideal for building AI agents.
3/2/2026
Show HN: Smart-commit-rs – A zero-dependency Git commit tool in Rust
Rust CLI tool for Git commits with LLM integration supporting multiple providers (Groq, OpenAI, Anthropic, Gemini). Follows Conventional Commits and Gitmoji standards.
3/2/2026
Ask HN: Self Sustaining Codebases
Discussion of self-sustaining codebases where agents autonomously fix bugs, add features, refactor code, and manage dependencies without human intervention.
3/2/2026
Show HN: Ccbridge – A CLI to Orchestrate Claude Code and Codex
Open-source CLI orchestrator for multi-agent workflows combining Claude Code and Codex with planning, execution, and review loops.
3/2/2026
Show HN: Self-hosted AI agent observability (OTel, Grafana, bash hooks)
Open-source tool for observability of AI agents using OpenTelemetry, Grafana, and bash hooks for monitoring and debugging.
3/2/2026
Show HN: Independent monitoring of AI API reliability
Independent monitoring of LLM API reliability comparing latency metrics across Google Gemini, Anthropic Claude, and OpenAI GPT models.
3/2/2026
Show HN: Valkey-powered semantic memory for Claude Code sessions
Valkey-powered semantic memory layer for Claude Code sessions enabling context retrieval without repeating information.
3/2/2026
Show HN: Workz – run 5 AI agents on parallel Git worktrees with one command
workz: Tool automating git worktree setup for multi-agent parallel execution with automatic dependency/env detection and fleet mode.
3/2/2026
We Audited 2,857 Agent Skills. 12% Were Malicious
Security audit of 2,857 Agent Skills found 12% contained malicious code. Analysis of supply chain risks in agent ecosystem.
3/2/2026
I built a new Terraform agentic editor and auditor
Grafos tool converts Terraform code to interactive visual graphs with AI assistance for compliance enforcement using natural language policies.
3/2/2026
Show HN: Gipity – AI cloud computer in the browser
Browser-based AI cloud computer called Gipity with persistent state, terminal interface, and programmable workspace. OpenClaw-inspired approach.
3/2/2026
Open-sourced an AI-native habit tracker where the LLM is the interface and coach
Open-source habit tracker using LLM as interface and coach. JSON-native, deterministic sprint-based engine designed for LLM-first workflows.
3/2/2026
Show HN: Lysium – cross-platform control plane for agentic software delivery
Lysium: control plane for agentic software delivery integrating GitHub and Devin API to parallelize agent work on issues/PRs.
3/2/2026
UE LLM Toolkit – Open-source plugin that lets LLMs read/write UE assets"
Open-source C++ plugin exposing 37 tools and 200+ operations via REST API, giving LLMs read/write access to Unreal Engine 5 projects.
3/2/2026
Show HN: Alpha of a Speech to Text iOS App (Feature Requests Welcome)
iOS speech-to-text app using whisper.cpp and Apple Speech models with source code available. Early-stage consumer app.
3/2/2026
Show HN: Tensor Spy: inspect NumPy and PyTorch tensors in the browser, no upload
Tensor Spy is a browser tool for inspecting NumPy and PyTorch tensors locally without uploading data, built using agentic coding.
3/2/2026
Show HN: Product Model – A structured grammar for bridging PRDs and code
Open-source Product Model: MDX-based grammar for product specifications with typed validation and JSON AST output bridging PRDs and code.
3/2/2026
Show HN: Engram – Give your terminal an eidetic memory with local AI
Local AI tool that logs terminal commands/output to SQLite and enables natural language queries. Uses Anthropic API optionally, fully private by default.
3/2/2026
Show HN: btrc – C, but better. Built with AI in just a few weeknights
btrc: statically-typed language transpiling to C11 built with AI assistance. Demonstrates AI-assisted language design with formal specifications.
3/2/2026
Ask HN: How do you monitor AI features in production?
Ask HN discussion on monitoring output quality of LLM-powered features in production when traditional error tracking doesn't apply.
3/2/2026
Show HN: Benchmarking the Keep memory system with LoCoMo
Keep is an open-source memory system for AI agents based on reflection and iteration. Benchmarked with LoCoMo evaluation framework.
3/2/2026
Who is done with code reviews?
Opinion piece discussing AI capabilities for code review versus code generation. Light on technical substance.
3/2/2026
TrustPlane – open-source AI trust control plane
Open-source AI trust control plane for LLM deployments with compliance documentation, audit trails, and automated human review escalation. Based on CognOS epistemic scoring.
3/2/2026
Claude Code NPM downloads up and50% in recent weeks
Title only: Claude Code NPM package downloads increased 50% recently.
3/2/2026
AgentPay: Budget controls and approval UI for autonomous AI agents
Title only: Budget controls and approval UI framework for autonomous AI agents.
3/2/2026
Vercel's AI Sales Agent: 10 SDRs Down to 1 in Six Weeks
Title only: Vercel deployed AI sales agent replacing 10 SDRs with 1 agent in six weeks.
3/2/2026
AgentCgroup: Understanding and Controlling OS Resources of AI Agents
Title only: Framework for monitoring and controlling OS resource usage by AI agents.
3/2/2026
An in-kernel machine-learning library
Proposal for integrating ML models into Linux kernel subsystems for configuration and logic optimization through data collection, training, and inference phases.
3/2/2026
Show HN: EvoAgents – Agents that evolve their own skills
EvoAgents is an open-source framework where multi-agent systems automatically improve their own prompts through LLM judges scoring failures and patching skills.
3/2/2026
Anthropic accuses Chinese AI labs of mining Claude
Anthropic accuses Chinese AI labs of using distillation technique on Claude to steal agentic reasoning, tool use, and coding capabilities.
3/2/2026
Hybrid Multi-Agent Congressional Simulator
Multi-agent system simulating 535 AI-powered Congress members debating and voting on bills following real legislative procedures.
3/2/2026
Show HN: Dbcli – Database CLI Built for AI Agents
dbcli: Database CLI tool optimized for AI agents to query, profile, and understand database schemas and relationships.