Releases·Tool Release
Show HN: KISS – A Rust agent harness inspired by Pi, with sub-agents, MCP, and browser WASM builds
A developer launched KISS on GitHub, a performant agent harness written in Rust and modeled on the Pi harness, created after frustration with Pi's performance. It targets practitioners who want an embeddable, multi-language agent runtime.
4 engagement1 sourceSat, Sep 26, 2026, 12:05 AM
AI Overview
- Developer racetozero built KISS, a Rust agent harness inspired by Pi's principles, citing unhappiness with Pi's performance
- KISS adds optional sub-agents, dynamic workflows, MCP, ACP, and WebMCP support
- The harness supports opt-in Jev for dynamic compaction and mid-turn reasoning/effort changes
- SDKs ship for Rust, Python, TypeScript, and WASM, plus JSONL and WebSocket RPC for other languages
- The WASM build runs the full agent loop in the browser without a server, and the harness embeds in other projects
- The author daily-drives KISS for work and personal use; performance benchmarks are documented on the GitHub repo
Related
Releases·Tool Release
Pi 1.0 released with MCP support included by default
Earendil releases Pi 1.0 with MCP support included by default
Earendil releases Pi 1.0 with MCP support included by default
546 engagement1 sourceredditFri, Oct 2, 2026, 12:09 AM
Releases·Agent Release
Aex opensources Brain, a Rust agent runtime with Wasm agent loops and stateful sessions
The Brain team open-sourced Brain, an MIT-licensed Rust runtime for persistent agent services, in an early preview announced on Reddit seeking feedback. Brain separates the agent decision loop, which compiles to Wasm, from the runtime that handles session state, model and tool I/O, streaming, cancellation, restart recovery, and journaling. It targets agents that have outgrown simple scripts, where infrastructure rather than the model loop is the hard part.
The Brain team open-sourced Brain, an MIT-licensed Rust runtime for persistent agent services, in an early preview announced on Reddit seeking feedback. Brain separates the agent decision loop, which compiles to Wasm, from the runtime that handles session state, model and tool I/O, streaming, cancellation, restart recovery, and journaling. It targets agents that have outgrown simple scripts, where infrastructure rather than the model loop is the hard part.
6 engagement2 sourcesredditWed, Sep 2, 2026, 10:44 AM
Releases·Tool Release
Dev releases Brain, an open-source minimal agent runtime pitched against LangChain and Mastra (<1ms session, ~25MB binary)
A developer open-sourced 'brain' (github.com/aexhq/brain), a minimal, extensible agent runtime inspired by Anthropic's managed agents engineering post and pi.dev. It targets production agent use cases the author found too rigid in LangChain and Mastra. For practitioners, it offers a lightweight alternative with sub-second cold starts and a small compiled footprint.
A developer open-sourced 'brain' (github.com/aexhq/brain), a minimal, extensible agent runtime inspired by Anthropic's managed agents engineering post and pi.dev. It targets production agent use cases the author found too rigid in LangChain and Mastra. For practitioners, it offers a lightweight alternative with sub-second cold starts and a small compiled footprint.
2 engagement1 sourceredditMon, Sep 7, 2026, 10:08 PM
Products·launched
Minimalist coding-agent harness compiled to WebAssembly (~130 KB)
h5i-agent is a coding-agent harness compiled into a WebAssembly module of about 130 KB, designed to be extremely small and portable. It runs the agent loop without requiring a large Python or Node.js runtime, making it suitable for lightweight or embedded environments. The harness likely uses LLMs to power the agent's decision-making, though the post focuses on the minimal footprint.
h5i-agent is a coding-agent harness compiled into a WebAssembly module of about 130 KB, designed to be extremely small and portable. It runs the agent loop without requiring a large Python or Node.js runtime, making it suitable for lightweight or embedded environments. The harness likely uses LLMs to power the agent's decision-making, though the post focuses on the minimal footprint.
6 engagement1 sourceredditThu, Aug 27, 2026, 04:29 PM
Products·launched
MCP server for scheduled news summaries via Claude
An MCP server that sits between a publishing backend and Claude, enabling scheduled news summarization. It collects news via RSS/Atom feeds throughout the day and generates updated article overviews hourly, triggered by scheduled tasks. Built with Claude Code to avoid direct API calls and control costs.
An MCP server that sits between a publishing backend and Claude, enabling scheduled news summarization. It collects news via RSS/Atom feeds throughout the day and generates updated article overviews hourly, triggered by scheduled tasks. Built with Claude Code to avoid direct API calls and control costs.
4 engagement1 sourceredditTue, Aug 18, 2026, 06:45 PM