The Drop #3: payments for indie SaaS, coding agents that hand off work, and a beginners' list

Hey builders. Here's what we added and published since The Drop #2 on Oct 26. Every tool description comes from the tool's listing or its own site.

New pages

• Stripe vs Polar vs Lemon Squeezy vs Paddle. Stripe, Polar, Lemon Squeezy, Paddle, Dodo Payments and Creem side by side: fees, who acts as merchant of record, and who files your sales tax, all from each provider's own pricing page.
• Best AI coding tools for beginners. Picks sorted by where you're starting: never coded, learning in an editor, or ready to hand work to an agent. Free plans are noted.

A newsletter we like

The Vibe Diff: a Monday email on what changed in AI coding tools that week, with a source on every item. https://the-vibe-diff.vercel.app

New tools

Coding agents that share the work

• Friday: a free, MIT-licensed harness where Claude Code, Codex, Cursor, Gemini CLI and Copilot split work, message each other and hand off when one hits a usage limit.
• ReplySlip: an open-source task inbox over MCP, so Claude Code on your laptop can hand a task to Codex on a server and the reply gets checked against a JSON Schema.
• Palomino: a GPU-rendered Rust workspace that runs Claude Code, Codex and Pi in side-by-side panes with terminals, browsers and files. Free, no account.
• Kora (runkora): ClearRoute's open-source control plane for running Claude Code, Codex and OpenCode on your own Kubernetes, with sandboxes, per-agent identity and budgets.
• SuperPlane: an Apache 2.0 platform that walks coding agents from ticket to merged pull request.
• DeepSeek Super-Agent: DeepSeek's MIT-licensed research prototype that watches a GitHub repo, takes requests from approved users and answers each one with a pull request.
• Spec-First: a Claude Code plugin that writes and reviews a spec, builds it in an isolated worktree and delivers a pull request through your checks.
• Toolmaker Agent: a free single-binary local workbench where you and your agents keep one shared record of what you're building.
• Entire: a code collaboration platform from former GitHub CEO Thomas Dohmke, built around Trails instead of pull requests.
• lever: GitHub's MIT-licensed Copilot plugin that turns a vague problem into a Loop Card with a metric and a success bar, so the agent can iterate on its own.
• Simlock: Callstack's open-source simulator pool, so each coding agent gets its own iOS or Android simulator or device.
• Flatbread: turns a Claude Code, Codex or Cursor session into one shareable page with prompts, failures, fixes and the diff, with secrets redacted locally.
• BuildGist: a Mac app that explains in plain English what changed after each Claude Code, Cursor or Codex run and what to check.
• CodeConnect: a free iPhone app for watching and steering the agents on your Mac or Linux box: approve commands, read diffs, open a terminal.
• Rivet: a free, open-source coding agent that runs on your Android phone, with your own key or a ChatGPT plan.
• Local Code: an open-source agent from Draw Things that runs models like DeepSeek 4.1 Flash fully on your Mac, no API keys.

Guardrails and observability

• safeagents: Hugging Face's open-source local proxy that checks every tool call between your coding agent and the model before it runs.
• Grim Prompter: open-source threat-modeling prompts; run /threat-model on your repo and get threats and fixes with file and line citations.
• Judgment Labs: captures full traces from long-running agents, runs custom judges to find repeat failures and lets you test fixes offline.
• telemetry.dev: OpenTelemetry tracing for model calls, tool steps and retrievals, with token cost, latency and MCP failure stats.
• Overmind: open-source platform that turns your agents' production traces into small models you own, with an OpenAI-compatible API.

Agents that do a job for you

• Whenever: build workflows, AI steps and agents from your coding agent and run them per run, with 100 free runs a month.
• BetterClaw: describe a job in chat, bring your own LLM key, and the agent runs it on a schedule in Slack, Telegram, Discord or Gmail.
• Organ: early-access agents that run a venture's departments on a schedule, with merges, outbound email and publishing waiting for your approval.
• Agent Desktop: Vercel Labs' open-source Rust tool that lets agents drive macOS, Windows and Linux apps through the accessibility tree instead of pixels.
• fastbrowse: an MIT-licensed browser agent that returns cited answers or schema-checked data, usable from Codex or Claude Code.
• UserAlertX: your bot sends one HTTP POST and you get an SMS or email, only to the phone number you verified.
• Pidgeon: EU-hosted email inboxes for agents, with threads, webhooks, IMAP and MCP.
• SlopIt: a blog your agent signs up for, writes and publishes; humans can't post directly.
• StoreRouter: one API and CLI for agents to search stores and parse product pages without scraping, $5 of free credit a month.
• Limma: one MCP server that reads your company's documents and data and answers your agent with proof.
• phpClaw: an MIT-licensed agent runtime for PHP with tools, memory and guards, for Laravel, Symfony or WordPress.
• Aura Autopilot: Ramen's open-source playtester that plays your Unreal, Unity or Godot game from plain-English goals and writes a bug report.

Taking payments

These six are the providers in this week's payments comparison.

• Polar: merchant of record for developers, with usage-based billing, credits, seats and trials.
• Lemon Squeezy: merchant of record for SaaS and digital products, with license keys, a storefront and affiliates.
• Paddle: merchant of record for SaaS and mobile apps, covering checkout, invoicing, tax and buyer support.
• Dodo Payments: merchant of record for software and AI products, with credit-based billing and usage metering.
• Creem: merchant of record that collects VAT, GST and sales tax in 50+ countries, with revenue splits built in.
• tiun: auth, checkout, a customer database, a customer portal and analytics in one backend, set up through an SDK or MCP.

Backend, hosting and local dev

• TeamOnly: deploy an internal tool you built with Claude Code or Cursor in one command; only coworkers with a company email can sign in.
• Adios: isolated AI workspaces, an MCP server for coding agents and full-stack hosting, with a free workspace.
• CloudHike: Postgres, storage, Redis, auth, email, jobs and 100+ AI models behind one API key.
• Sectr: deploy an agent from OpenAI Agents SDK, LangGraph or Pydantic AI as a serverless backend with durable sessions.
• Groundmail: transactional email your coding agent sets up through MCP or the CLI.
• Flisk: an open-source CLI that runs your website, API, database and agent locally with one `flisk up`.
• Expo Agent CLI: Expo's experimental CLI that gives coding agents a structured way to create, run and debug Expo apps.
• Lasm: an open-source toolchain for building web apps front to back in Lean and formally verifying them.

UI and design

• Clonecode: paste a URL and get its colors, fonts and spacing as DESIGN.md and tokens.json, plus prompts for Claude Code, Cursor, Lovable, v0 and Bolt.
• GradientDeck: an MIT-licensed Agent Skill that gives coding agents design judgment, for Claude Code, Codex, Cursor and Copilot.
• Wingo UI: 350+ animated React 19 and Tailwind v4 components (75 free), with an MCP server for your agent.
• Avara UI: 190 free shadcn/ui components with GSAP and Motion animations, installable by CLI or MCP.
• Weasel: 874 animated React SVG icons that ship with a coding-agent skill.
• VibeBoot: Next.js 16 starters with Auth.js, Stripe and Prisma wired in, plus Cursor rules and an AGENTS.md.

Models

• Mistral AI: Mistral's Vibe chat assistant and coding agent (CLI, VS Code, JetBrains or Zed), plus its API in Mistral Studio.
• Particle: Particle-1, a coding model Particle says uses about 45% fewer tokens than comparable models, on its API and Vercel AI Gateway.
• llmrun: tells you which open-weight models your GPU can run, with VRAM needs and quantization options.

Data, notes and small APIs

• Hob: a free, MIT-licensed desktop app that edits files, fills PDFs and runs code with open-weight models on your machine, offline.
• static.md: shared markdown notes and whiteboards that Claude, ChatGPT, Cursor and VS Code can read and write over MCP.
• BulkTranscripts: pulls every transcript from a YouTube channel or playlist, up to 1,000 videos a job, in AI-ready formats.
• Registrum: a Companies House API and hosted MCP, so you can ask Claude about any UK company.
• ReadGZH: turns public WeChat Official Account articles into AI-readable HTML or Markdown, by site, API or MCP.
• FaviconDL: grab a site's favicon through a web tool, a free keyless API or a local MCP.

App store growth

• Winch: a Mac app that lets your agent handle App Store and Google Play optimization over MCP, with nothing submitted until you review it.
• ASOClimb: keyword research and daily rank tracking for indie apps on the App Store and Google Play.

Video

• Arsaze: an AI video editor with Tab-complete suggestions, agent edits, auto captions and an MCP server, free plan included.
• Overtone: an open-source CLI and video.js plugin that cuts captions cleanly for 16:9, 9:16 and other shapes.

Added this morning

• Clawdline: a local daemon and web console for macOS and Linux that finds your Claude Code and Codex sessions, shows which need you, and lets one session hand work to another.
• Skep: an open-source CLI for live cross-device sessions between AI coding agents, using join codes, with packaged Skills and role briefs.
• rufous: an Apache-2.0 agent harness for 2B to 12B models on your own machine that records every tool call and checks the model's report against it.
• Kilasec: see what every AI agent calls, cap runaway spend, hold risky actions for a human and mask secrets and PII, deployed as a single container.
• nonce: an open-source layer between AI apps and your email, calendar, files and Slack, where each request gets a short-lived pass to just what it needs.
• Enclyra: an offline AI system for Mac built around a local 3-4B model that does multi-step tasks on 8 GB machines, with a free core.
• OpenMarkdown: a local-first Markdown editor built for you and your AI agent to edit the same docs together.
• Corollary: a free, open-source Overleaf-style LaTeX editor for Mac whose AI runs on-device, offline, with no account.
• Storytella: turns articles and websites into natural speech entirely on your device with WebGPU, no account.
• publishub: shows the exact second viewers drop off your Reels, TikToks and Shorts, why they leave and what to cut.

Browse more

The full directory is at getstackdrop.com/tools, and every comparison lives at getstackdrop.com/compare. Built something? Submit it.