Developer Corner

An open benchmark tests how different search providers affect AI coding agents on research and coding tasks.

An open-source cloud drive that stores files in your Cloudflare R2 bucket and serves requested file ranges to Mac, iPhone, and browser clients.

An open, three-character convention for disclosing who maintains a code project, how much of its code was AI-written, and how people check that work.

A solo developer’s portfolio pairs browser-playable games with game engines, scripting languages, and development tools.

A monitoring service for certificates, domains, email-security records, and other assets that can expire or break.

loopl is an on-device iPhone agent app and Swift SDK, designed to run models and native tools without an account or network connection.

A Northeastern student club designs, builds, and launches small satellites, with several missions and supporting hardware projects underway.

A live index maps personal AI machines, separating vendor claims from measured, observed, inferred, unverified, and failed results.

An open-source pull-request review assistant that uses Claude Code to draft findings from a diff, then leaves editing and posting to the developer.

RSSgate turns RSS feeds and selected web pages into a locally hosted reading stream, using an LLM to transcribe articles and strip ads and boilerplate.

A browser-based checker compares an AI endpoint’s answers to low-stakes prompts against a trusted reference, looking for signs that the endpoint may be serving a different model.

Walkabit lets you dictate tasks to Claude Code from an Apple Watch and send them to a Mac through iCloud.

cint is an experimental C-family language built around checked integer arithmetic and deterministic output across platforms.

Leo presents an Italian-language programming environment, with examples, a browser trial, and downloads advertised for Windows and Android.

A free macOS app that says it tracks coding-agent sessions, editor windows, websites, and terminals by repo or worktree, without replacing the tools themselves.

OpenLFCP is an open protocol and an Obsidian plugin for syncing a shared task between people’s private Markdown notes.

A browser-based formatter, validator, and converter for code, data, and configuration formats.

An open-source canvas for editing real components in a React app, with changes applied to the codebase and reviewable through a Git workflow.

Agents Discuss describes a relay that lets two AI agents exchange text messages, with people coordinating the connection through a short-lived join code.

A Mac app that collects Claude Code, Codex, and Cursor chats, masks them locally, and makes the archive searchable through MCP.

A Mac and Linux tool that checks a code repository for architecture changes and drafts diagram updates for human approval.

A free, open-source Windows app for routing and processing audio, applying VST effects, and recording separate signal branches.

boxdbox describes a 3D office in your browser where the desktop runs on a real Linux cloud machine.

Chickadee is a runner platform for GitHub Actions, offering disposable Linux virtual machines for individual jobs, with hosted beta access and a self-hosting option.

A set of specialized electromagnetic simulation studios for moving objects, antennas, and microwave devices.

A live register and mint lookup for Solana tokens with Token-2022 confidential transfers enabled, including their supply and controlling authorities.

Firmpath says it matches vulnerability advisories against shipped firmware builds, then prepares customer notices, VEX statements, and disclosure records.

A filesystem for people and AI agents to edit shared plain files, with section-level turn-taking and history controls.

Leon brings coding agents into a native desktop workspace, organizing sessions across projects, worktrees, and local or remote machines.

A free, in-person Madrid meetup for people developing with AI, built around three talks and time to meet other developers.

An interactive explainer and playground for making repeated tiny payments off-chain, then settling their cumulative amount on Solana.

An open-source engine for organizing notes into typed facts, then showing what is known or missing for a chosen goal.

A paid companion for monitoring Claude Code sessions across computers and responding to permission prompts, questions, and completion alerts from a phone.

Mazapan presents an Arch-based desktop for developers, with coding-agent sessions, system updates, and theming built into the desktop.

modscode maps common mod behaviors to examples drawn from public Claude Code repositories, with code snippets organized by what a builder wants to make.

Naskah is a programming language with Indonesian syntax and a browser-based playground.

A macOS app for proxy routing, DNS rules, and network diagnostics, currently offered in system-proxy mode.

A society for amateur radio operators, engineers, educators, and others working with satellite and microwave communications.

OpenJuvenile is drafting a shared product-data format for baby and toy goods, intended to let brands describe a product once and reuse that information across retail channels.

A five-stage pipeline that aims to turn public vulnerability intelligence into plain-language email digests for small organizations.

A service lets AI agents create phone-call tasks for human callers, then receive the recorded outcomes through an API.

PMARC Labs is developing small actuator modules for robotic hands, combining powered motion with built-in torque and position sensing.

Pomme is a command-line tool for creating and automating headless macOS virtual machines on Apple silicon.

An open-source MCP server that lets AI clients access Google Play release status, reviews, and app reports using your own account.

A free, open-source desktop project journal that routes AI-agent summaries through MCP into Markdown files on your disk.

A small local Markdown editor for reviewing files changed by AI agents, with a terminal, live preview, and focused diff view.

Thiral is a research programme proposing modular robot swarms that assemble large structures in orbit and on the Moon.

Boardless is an in-development tool for checking a project’s schematic, enclosure, and firmware against one another.

A platform for running AI agents in Linux VMs and collecting their output files through a CLI, HTTP API, or SDK.

VhyxChart turns text-defined flow, sequence, state, and array diagrams into animated examples for documentation, GitHub, VS Code, and React.

An open-source, local-first tool that records AI coding-agent sessions and lets you rewind project files to earlier steps.

Livery presents a tool for turning website references into design kits that coding agents can use to restyle an app.

A science platform for people and AI agents to submit claims, test reproductions, and preserve evidence in an auditable ledger.

Tanod lists security checks and utilities that agents can call over HTTP or MCP, from Solidity scans to DNS inspection and chain data.

TAVIS is a simulation benchmark for evaluating how robots use movable cameras while learning manipulation tasks from demonstrations.

A prototype observability tool for reconstructing where a physical-AI task went wrong.

TurnGraph presents an API for giving voice agents step-by-step routes through company phone menus.

Graveyard proposes a version-pinned library where agents can check for verified failure cases before acting.

A browser-based scanner reads an npm lockfile and lists the people who can publish code into a project’s dependencies.

A Portuguese-language API for querying and monitoring Brazilian company-registry data, with each response identifying its source month and reported coverage.

A browser extension for filling web forms with generated test data, aimed at QA engineers and developers.

A Mac utility for changing an iPhone’s reported location while testing apps that depend on GPS.

A Tokyo pilot lets AI agents request human checks of real-world details, then returns evidence and a verified result through an API or MCP.

A software-adapted TRIZ CLI and library helps coding agents analyze conflicting requirements and return evidence-bound strategy suggestions for a human to review.

A web-capture API that offers screenshots, PDFs, Markdown, and AI-extracted JSON from public pages.

The proposed GitHub check asks unfamiliar contributors and AI agents for a refundable USDC deposit before their pull requests are reviewed.

A Windows app scans local Blender project files to show which projects reference a selected texture, model, or other asset.

Sitr presents an API layer that masks personal data in prompts before forwarding them to an AI model, then restores the values in the reply.

Hive is an MCP server that helps an AI agent choose a tool, call it, and assess whether its result answered the request.

A catalog of 34 data endpoints that AI agents can call and pay for in USDC through the x402 protocol.

Drift is a local browser-testing tool that records a narrated session and assembles issues with screenshots, logs, and reproduction steps.

HumanAsker presents an MCP server that lets AI agents ask people for opinions and receive answers they can use mid-task.

Kiste offers Linux microVMs that can be stopped with their memory intact, then resumed with processes still running.

Meter describes a proxy for tracking Claude Code usage by developer and repository, with security flags and controls for risky actions.

An interactive introduction argues that verified physical execution could provide a permissionless basis for consensus.

PhoneFleet offers dedicated physical Android phones hosted in racks and controlled through a browser, REST API, ADB, or MCP.

A California DROP compliance API for data brokers that checks records or CSV uploads against the state deletion list and returns matches marked for deletion.

Sunset scans a repository for risk from scheduled OpenAI model shutdowns, offering free severity counts and paid reports with file-level migration details.

An interactive observatory for website bot traffic, mapping crawlers, scanners, and other visitors in a reserve-themed dashboard.

A custom-development offering that combines sensors, algorithms and low-power chips into devices designed to make decisions offline.

A browser-based inspector checks supported IAMF metadata in standalone IAMF files or MP4s containing IAMF audio.

RunLite lets MCP-compatible AI tools deploy static sites, add forms, and route submissions to an inbox.

A self-hosted risk layer that checks bot and AI-agent orders against stop, leverage, position, and account-loss rules.

Sturdy SIM offers a data-only SIM for connected devices, with access to more than 200 cellular networks, automatic failover and a flat-price pitch.

A messaging interface for people and AI coding agents, with inbox, contacts, groups, drafts, and MCP/API access.

An early-stage project plans to re-derive and rerun calculations from astrophysics and cosmology papers, publishing reviews and code on GitHub.