MCP

Show HN: Omnara – Open-source alternative to Claude Managed Agents

Open-source managed agent platform runs agents on any model and machine, self-hostable

Launch HN: Alter (YC S25) – Access control for multi-user agents

Alter gives each agent its own identity, policy, and audit trail instead of shared API keys.

Harness Engineering 101: The LLM is the brain, the harness is the body

An agent is a loop managing a JSON array; 16 chapters build one from scratch

If you don't run your own LLM proxy, how do you know whats going on?

Running your own LLM proxy shows which keys, models, and cache misses drive your usage.

AI-assisted PCB design for KiCad 10

Konnect exposes KiCad to Claude via MCP tools in one Rust binary, cutting listing context from 23K to 2K tokens

Docker Agent

Define multi-agent setups in YAML and run them with docker agent

Jevmem – automatic project memory for Claude Code, built on Jev

Saving chat decisions to JEVMEM.md raised Claude's adherence from 28/72 to 66/72 sessions.

Zero-Trust Authorization and Discovery for Enterprise MCP

Combining tool-list filtering with call-time checks cut forbidden-tool attempts from 21.1% to 0% in MCP.

Datamimic – don't let your coding agent invent its own test world

Pair a seeded synthetic data engine with your coding agent to get reproducible test data.

Reverse-Engineering Claude Web's MicroVM: Uncovering Anthropic's Hidden Antspace

Poking at a Claude Code web session exposed Anthropic's undisclosed deploy platform, Antspace

OpenAI Agents API

OpenAI launched a managed agents API handling sessions, context compaction, and recovery.

No-Box Vulnerability Analysis: Description-only Detection of Indirect Prompt Injection Vulnerabilities in MCP Servers

Feeding only MCP tool descriptions to an LLM found 94 of 95 confirmed prompt-injection-vulnerable tools.

Show HN: Engrim – A universal, local-first SQLite memory engine for AI CLIs

Swapping transcript replay for a 4,000-char memory pack kept decisions alive across /clear and tools

OKF Agent Memory – Git-native persistent memory for AI coding agents

Git-markdown memory with in-memory BM25 search claims 80% fewer tokens and sub-300µs lookups

AI Engineer Notebooks – free, framework-free RAG/agents/evals on Colab

Free Colab notebooks teach RAG, agents and evals from raw API calls on Groq, no frameworks

CEO fired developers to make room for AI. Developers create open source AI CEO

A multi-agent open-source 'executive' fuses nine specialists into one voice with caching and memory.

WebMCP: Teaching Your Website to Talk to AI Agents

Let sites declare tools and agents call them without guessing at the DOM

Beyond the Mandate: A Systematic Security Analysis of the Agent Payments Protocol (AP2)

Catalogs 48 AP2 threats: validly signed mandates can still carry unintended payments

Agent Is Not the Model

Splitting an agent into model, inference service, and harness shows which layer to fix.

fx :Tiny, open, native coding agent.

Vercel Labs released fx, a 6.66MiB single-binary coding agent written in Zig.

Launch HN: Speko (YC S26) – OpenRouter for Voice AI

Speko routes STT, LLM and TTS through one API using per-language benchmarks

Show HN: MCP Memory – Fast Agent Memory Using Google's OKF and SQLite FTS5

MCP server gives agents long-term memory via OKF markdown files indexed with SQLite FTS5

Docker Sandboxes – Disposable, isolated sandboxes for AI agents

Running agents in microVMs lets you skip permission prompts while keeping the host safe

Show HN: The Channels SDK – Bring Any Agent to Any Channel (Slack, MS Teams)

Bring any AG-UI agent to Slack and Teams with native interactive UI, keeping your agent code

Cloudflare OS: an open platform for agents, apps, and work

Cloudflare open-sourced an agent workspace where agents start with zero access and policy follows what they've read

qm – Multiplayer agent harness for work

Per-person and per-room isolated sandboxes let a team share one agent across Slack and web

Agent-Manager: A Tmux TUI for Running Claude Code, Codex and OpenCode

A tmux TUI that lists coding agents with live status, quick prompts, and diff review in one view

Show HN: OneCLI – OSS credential gateway that keeps secrets out of AI agents

A gateway injects credentials into requests so AI agents never see the secrets.

ANSI escape injection in MCP servers: Hidden from humans, visible to AI

ANSI escape codes hide instructions from human reviewers but the model reads them in full

Agent-talk: Enabling coding agents to work together

agent-talk lets coding agents message each other directly over end-to-end encryption

Grok Build is open source

SpaceXAI open-sourced its Rust terminal coding agent CLI/TUI under Apache-2.0.

Open-source memory for coding agents, synced over SSH

Indexing on-disk session logs lets coding agents recall fixes they already found

Designing APIs for Agents

Dropping defaults, using specific field names, and writing precise errors help agents misuse your API less.

Show HN: FableCut – A browser video editor AI agents can drive (zero deps)

A zero-dependency browser video editor whose timeline is one JSON file that AI agents can edit live.

Show HN: Microsoft releases Flint, a visualization language for AI agents

Microsoft released Flint, a visualization language for the AI era.

Show HN: Rowboat – Open-source, local-first alternative to Claude Desktop

Local-first open-source AI assistant that builds a Markdown knowledge graph from your email and meetings.

Claude-real-video － any LLM can watch a video

Scene-aware frame extraction plus transcripts lets any LLM 'watch' video with fewer frames than 1fps sampling.

OpenWiki: CLI that writes and maintains agent documentation for your codebase

CLI turns your codebase into a linked Markdown wiki for agents and updates only what changed.

Show HN: OpenKnowledge – open source AI-first alternative to Obsidian/Notion

An open-source markdown editor that connects Claude and Codex to your notes via MCP

RubyLLM: A Ruby framework for all major AI providers

One Ruby API for 19 AI providers, with tools, agents, evals, and Rails integration

We built a persistent agent memory layer on Elasticsearch with 0.89 recall

Splitting agent memory into three indices with hybrid search reached R@10 0.89

HyperTool: Beyond Step-Wise Tool Calls for Tool-Augmented Agents

Folding tool call chains into one code block lifted Qwen3-8B accuracy from 9.93% to 33.33%.

MTG Bench: Testing how well LLMs can play Magic

Without a rules engine, gpt-5.5 medium led an LLM Magic turn benchmark at 95.4

Apache Burr: Build reliable AI agents and applications

Define actions and transitions in Python and get state persistence, UI debugging and replay for agents

Ask HN: What are tools you have made for yourself since the advent of AI?

HN developers share personal tools built with AI: sandboxes, MCP, QA automation

Collaborative Human-Agent Protocol (CHAP)

Logging human overrides as structured diff-plus-rationale events gives agents an auditable shared workspace.

Silurus/ooxml: Pixel-faithful Office documents, rendered in the browser

A Rust+WASM Office viewer on Canvas, built entirely by AI coding agents

Show HN: Paseo – Beautiful open-source coding agent interface

Open-source app runs Claude Code, Codex and other coding agents in parallel worktrees

Show HN: Ktx – Open-source executable context layer for data agents

ktx gives data agents approved metric definitions and warehouse context, so they query more accurately.

Multi-Agent LLM System for Automated Vulnerability Discovery and Reproduction

Adding fuzzer reproduction to LLM agents found 90% of AIxCC vulnerabilities

HarnessAPI: A Skill-First Framework for Unified Streaming APIs and MCP Tools

One skill folder auto-generates HTTP and MCP endpoints, cutting framework boilerplate by 74%.

Launch HN: Runtime (YC P26) – Sandboxed coding agents for everyone on a team

Payment teams run approval-gated AI agents in their own cloud with full audit logs.

Show HN: Semble – Code search for agents that uses 98% fewer tokens than grep

A CPU-only code search tool returns just snippets, cutting tokens about 99% versus grep+read.

How Claude Code works in large codebases

Layering CLAUDE.md, LSP, and hooks makes Claude Code work well on large codebases.

Show HN: Statewright – Visual state machines that make AI agents reliable

Per-phase tool limits took two local models from 2/10 to 10/10 on a small SWE-bench subset.

Principles for agent-native CLIs

10 principles for CLIs agents can use well; HN pushes back that it's just good CLI design

Agent-harness-kit scaffolding for multi-agent workflows (MCP, provider-agnostic)

An open-source TypeScript kit for scaffolding multi-agent workflows; details are sparse.

Show HN: Airbyte Agents – context for agents across multiple data sources

Airbyte launched an agent data-context product and claims token savings over vendor MCPs in its benchmark.

Show HN: Mljar Studio – local AI data analyst that saves analysis as notebooks

Ask in plain English; local Python runs and is saved as a reproducible notebook

Show HN: Filling PDF forms with AI using client-side tool calling

SimplePDF Copilot demo lets you fill and understand PDF forms by chatting.