We check every product on this page once a day and record whether it still answers.
An autonomous AI agent with its own cloud computer that finishes multi-step research, coding, or creative tasks unattended, for people who want delegated work done while they're away.
Works offline
What to know▼
How it works
An AI agent with its own cloud computer (browser, terminal, files, code) that takes a task and returns finished work — research, websites, spreadsheets, decks, audits, or tested code — with sourced evidence.
What's different
Can fan work out to parallel specialist agents and run on a schedule, and can be embedded via an OpenAI-compatible API or used in real repositories through Clark Code.
Best for
People who want to hand off multi-step research, coding, or creative tasks and come back to completed work.
We check every product on this page once a day and record whether it still answers.
A native macOS AI workspace with local-first chat, writing assistance, and browser automation, extensible by developers via an MCP registry.
Watch out · macOS only.
Works offline
What to know▼
How it works
A native macOS AI workspace with a multi-provider chat interface, local-first data handling, system-wide writing assistance, browser automation, file creation, and study tools that generate linked quizzes and flashcards.
What's different
Developers can extend it via an MCP registry, @-tools, and Prism Hosted capabilities.
We check every product on this page once a day and record whether it still answers.
Describes an agent in plain English and it builds a deterministic version that runs locally on your own hardware instead of a metered cloud API, for privacy-conscious teams that can't send sensitive work to the cloud.
Works offline
What to know▼
How it works
You describe an agent in one plain sentence, and Avery builds a deterministic version of it that runs on your own hardware rather than a metered public cloud, with no code required.
What's different
Runs locally and deterministically instead of routing through a per-token cloud API, aimed at avoiding cost, privacy, and lock-in problems of typical agentic AI projects.
Best for
Teams that want private, auditable automation they own outright, without sending sensitive work to a cloud API.
We check every product on this page once a day and record whether it still answers.
A debugger for AI agent runs that shows exactly what context each model call saw, for developers troubleshooting agents built with Claude Code, Codex, or Cursor.
Works offlineOpen source
What to know▼
How it works
Captures every step of an agent run, compares two runs, and can fork from any step to test changes
What's different
Local-first and open source, so traces never leave your machine
We check every product on this page once a day and record whether it still answers.
A single-binary CI test runner for LLM apps and agents that defines test cases in YAML and replays model or judge calls offline at no cost.
Works offline
What to know▼
How it works
Defines test cases and scorers in YAML, runs local targets on every pull request via REST or shell, with baselines, trials, model comparisons, and OTel traces.
Pricing
Replays offline for $0.
Best for
Teams that need CI regression testing for LLM apps and agents.
We check every product on this page once a day and record whether it still answers.
Local-first multi-agent runtime binary usable from terminal, scripts, or HTTP, with editable agent "constitutions" and tool allowlists.
Works offline
What to know▼
How it works
A single native binary agent runtime usable from the terminal, scripts, or over HTTP, with agents living in editable "constitutions" that enforce hard tool allowlists.
What's different
Local-first with network use optional.
Best for
Developers building multi-agent workflows who want local control over tool access.
We check every product on this page once a day and record whether it still answers.
Open-source, local-first AI system that coordinates multiple agents to plan, build, verify, and ship real software.
Watch out · Local-first operation.
Works offlineOpen source
What to know▼
How it works
Coordinates specialized AI agents to plan, build, verify, recover, and ship software, with persistent memory, browser evidence, and real deployment integrations.
We check every product on this page once a day and record whether it still answers.
Build AI agents entirely in the browser using local open-weight models over WebGPU, no install, no API key, no server, then export the whole thing as a single standalone HTML file. A developer tool for writing agent logic in Python.
Nothing to installWorks offlineCan export my data
What to know▼
How it works
Lets you build AI agents that run entirely in the browser: open models (Llama, Qwen, Gemma, Phi, DeepSeek) load locally via WebGPU, agent logic is written in Python running through Pyodide, and the whole thing exports as a single standalone HTML file.
What's different
No install, server or API key needed for the local models — prompts and data never leave the tab — and you can optionally bring your own OpenAI/Claude key, encrypted client-side.
Pricing
Free to start.
Best for
Developers who want to write and distribute browser-based AI agents without server infrastructure.
We check every product on this page once a day and record whether it still answers.
Runs AI agent chains, workflows and cron jobs locally on your own machine — local-first, private, bring your own API key. Free download for Linux, Windows and macOS.
We check every product on this page once a day and record whether it still answers.
A local-first persistent task-context store with an API that AI agents (and humans) can read and write to, so agent workflows keep project context across sessions.
