Harnesses: The program you actually run. It holds the conversation, calls the model, and edits your files, the agent itself rather than something you add to one.
The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery
The AI Scientist from Sakana AI is an experiment in automating the research loop: generating ideas, writing the code to test them, running the experiments, and drafting a paper about what happened. It is interesting less as something to use and more as something to look at, because it makes concrete a question everyone in the field is arguing about, which is how much of research is mechanical and how much is not. The output is real code and real drafts, and the honest reading of the results is that the pipeline works and the judgement of what is worth studying is still the hard part. Worth reading if you follow where agents are heading rather than looking for a tool to put to work tomorrow.
Assemble, configure, and deploy autonomous AI Agents in your browser.
AgentGPT lets you give an AI agent a goal in a browser tab and watch it break that goal into tasks and work through them, rather than answering one message at a time. You type something like research three competitors and summarise their pricing pages, and it plans the steps, runs them, and reports back. It is an open source project from Reworkd, and you can either use the hosted version or run the whole thing yourself. Either way it works by calling a language model on your behalf, so you supply an API key and the model provider bills you for what the agent uses. The appeal is seeing the planning happen in front of you. The catch is the same one every autonomous agent has: it is confident, it does not always know when it is stuck, and an open ended goal can run up a bill while going nowhere useful.
AutoGPT is the vision of accessible AI for everyone, to use and to build on. Our mission is to provide the tools, so that you can focus on what matters.
AutoGPT is the project that made autonomous agents a mainstream idea. You give it an objective, and it writes itself a task list, works through it using tools like web search and file access, and keeps going until it decides the objective is met. It is open source and has grown from the original script into a larger platform for building and running agents. It matters partly for what it does and partly for what it taught everyone: the demos that go well are the ones with a clear finish line, and the failures are almost always an agent that cannot tell it has gone wrong. If you want to understand how modern coding and research agents came to be shaped the way they are, this is the codebase that started the conversation.
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
Written mainly in JavaScript. Released under the MIT licence.
Fabric is an open-source framework for augmenting humans using AI. It provides a modular system for solving specific problems using a crowdsourced set of AI prompts that can be used anywhere.
Fabric is a command line tool built around the idea that most useful prompting is repeatable. Instead of writing a fresh prompt each time, you keep a library of patterns, small named prompts for jobs like summarising an article, extracting the argument from an essay, or turning notes into a structure, and you pipe text through them from the terminal. Anything that can produce text on standard input can feed it, which makes it easy to wire into scripts and existing habits. It is open source and the pattern library is public, so you can read exactly what each one asks the model to do and change it. It suits people who already live in a terminal; if you do not, the same ideas exist in friendlier packaging elsewhere.
Compiles a demonstrated GUI task into a program that reports VERIFIED only if an independent check agrees. pip install openadapt; openadapt flow tutorial. Compiler lives in openadapt-flow.
Written mainly in Python. Released under the MIT licence.
Research and development (R&D) is crucial for the enhancement of industrial productivity, especially in the AI era, where the core aspects of R&D are mainly focused on data and models. We are committed to automating these high-value generic R&D processes through R&D-Agent, which lets AI drive data-driven AI. https://aka.ms/RD-Agent-Tech-Report
Written mainly in Python. Released under the MIT licence.
SWE-agent takes a GitHub issue and tries to automatically fix it, using your LM of choice. It can also be employed for offensive cybersecurity or competitive coding challenges. [NeurIPS 2024]
Written mainly in Python. Released under the MIT licence.
< > SuperAGI - A dev-first open source autonomous AI agent framework. Enabling developers to build, manage & run useful autonomous agents quickly and reliably.
Written mainly in Python. Released under the MIT licence.
Graphs that teach > graphs that impress. Turn any code into an interactive knowledge graph you can explore, search, and ask questions about. Works with Claude Code, Codex, Cursor, Copilot, Gemini CLI, and more.
Written mainly in TypeScript. Released under the MIT licence.
Prompt, run, edit, and deploy full-stack web applications. -- bolt.new -- Help Center: https://support.bolt.new/ -- Community Support: https://discord.com/invite/stackblitz
Written mainly in TypeScript. Released under the MIT licence.
A cross-platform desktop All-in-One assistant for Claude Code, Codex, OpenCode, OpenClaw, Grok Build & Hermes Agent. Only official website: ccswitch.io
Written mainly in Rust. Released under the MIT licence.
Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflows - all through natural language commands.
Claude Engineer is an interactive command-line interface (CLI) that leverages the power of Anthropic's Claude-3.5-Sonnet model to assist with software development tasks.This framework enables Claude to generate and manage its own tools, continuously expanding its capabilities through conversation. Available both as a CLI and a modern web interface
Persistent Context Across Sessions for Every Agent to Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, OpenCode + More
Written mainly in TypeScript. Released under the Apache-2.0 licence.
Pre-indexed code knowledge graph, auto syncs on code changes, for Claude Code, Codex, Gemini, Cursor, OpenCode, AntiGravity, Kiro, CoPilot, and Hermes Agent, fewer tokens, fewer tool calls, 100% local
Written mainly in C. Released under the MIT licence.
Your agent in your terminal, equipped with local tools: writes code, uses the terminal, browses the web. Make your own persistent autonomous agent on top!
Written mainly in Python. Released under the MIT licence.
Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph. A /graphify skill for Claude Code, Cursor, Codex, and Gemini CLI: local deterministic AST parsing, every edge explained, no vector store.
Written mainly in Python. Released under the Apache-2.0 licence.
A powerful GUI app and Toolkit for Claude Code - Create custom agents, manage interactive Claude Code sessions, run secure background agents, and more.
Written mainly in TypeScript. Released under the AGPL-3.0 licence.
