Showing 110 agents
Adala
Autonomous Data Labeling Agent framework offering reliable performance through ground truth validation. Provides controllable output with flexible constraints, specializes in custom data processing workflows, features autonomous learning via observation and reflection, and includes an extensible runtime adaptable to diverse requirements without a steep learning curve.
Agent4Rec
Recommender system simulator powered by 1,000 LLM-based generative agents. Each agent embodies distinct social traits and preferences derived from the MovieLens-1M dataset, interacting with personalized movie recommendations through actions like watching, rating, evaluating, and providing feedback in a page-by-page interface.
AgentForge
Low-code framework for rapid development and iteration of AI-powered autonomous agents and cognitive architectures. Supports multiple LLM providers including OpenAI, Anthropic Claude, and local Oobabooga models. Features customizable memory management, on-the-fly prompt editing, and database-agnostic architecture with seamless ChromaDB integration.
AgentGPT
Browser-based no-code platform for deploying autonomous GPT agents. Users assign goals, observe thinking processes, and watch execution plans unfold. Leverages OpenAI functions, supports gpt-3.5-16k with extended context, and integrates Pinecone and pg_vector databases. Built on NextJS/TypeScript frontend with FastAPI/Python backend.
AgentPilot
Desktop application for building, managing, and conversing with AI agents. Seamlessly integrates with Open Interpreter for code execution and MemGPT for advanced memory management. Features multi-agent group chat functionality enabling collaborative conversations between multiple agents.
Agents
Comprehensive library for constructing language agents with dual-memory architecture combining VectorDB-powered long-term semantic memory and LLM-maintained short-term working memory. Features tool usage via function-calling, web navigation through search engines, dynamic multi-agent communication controlled by LLM-based coordination, human-agent interaction support, and SOP (Standard Operation Process) symbolic control for fine-grained workflow customization.
AgentVerse
Platform for assembling multiple agents to collaboratively tackle complex tasks through coordinated teamwork. Enables creation of custom environments where multiple agents can be observed or engaged in interactive simulations. Supports both task-solving workflows and simulation scenarios with emergent multi-agent behaviors.
AI Legion
TypeScript-based autonomous agent platform enabling multiple agents to collaborate on task completion. Provides framework infrastructure for coordinated multi-agent operations with console-based direct message interactions between agents and users.
Aider
Command-line pair programming tool for GPT-3.5/GPT-4 integration with local git repositories. Supports both new projects and existing codebases, enabling fluid switching between AI-assisted editing and manual code modifications. Automatically commits changes from both AI and developer with meaningful commit messages, excelling with larger pre-existing codebases.
AIlice
Chatbot agent that independently plans and executes tasks described in natural language by dynamically constructing a hierarchical agents calling tree. Features inter-agent interaction mechanisms for fault tolerance and automatic building of external interaction modules for self-expansion capabilities.
AutoGen
Microsoft framework for developing LLM applications using multiple conversational agents that collaborate to solve tasks and interact with humans. Simplifies complex LLM workflows with enhanced automation and optimization. Offers easy performance tuning, API unification, caching, and supports advanced patterns including error handling, multi-config inference, and context programming.
AutoGPT
Experimental open-source project making GPT-4 fully autonomous with over 140k GitHub stars. Chains LLM thoughts to autonomously achieve user-defined goals. Features internet access for searches and information gathering, long-term and short-term memory management, command execution including Google Search, website browsing, file operations, Python script execution, access to popular platforms, GPT-3.5 file summarization, and plugin extensibility.
Automata
GPT-4 powered code generation system that leverages entire project context to execute development instructions seamlessly. Aims to evolve into fully autonomous self-programming AI system with integration for all agent platforms and LLM providers. Utilizes novel SymbolRank code search algorithm for superior coding intelligence with modular, configurable design and minimal external dependencies.
AutoPR
Automated pull request agent activated by adding an AutoPR label to GitHub issues. Autonomously plans fixes, writes implementation code, pushes branches, and opens pull requests for review. Streamlines the entire code contribution workflow from issue identification to PR submission.
Autonomous HR Chatbot
GPT-3.5 powered prototype enterprise HR assistant that autonomously answers human resources queries using specialized tools. Current toolkit includes Timekeeping Policy database, Employee Data access, and Calculator functionality with additional tools in development pipeline.
BabyAGI
Simplified version of Task-Driven Autonomous Agent that creates tasks based on previous results and predefined objectives. Leverages OpenAI GPT-4, Pinecone vector search, and LangChain framework. Maintains task list for management and prioritization, autonomously generating new tasks from completed results and dynamically reprioritizing, showcasing AI language model adaptability. Default model is GPT-3.5-turbo.
BabyBeeAGI
Advanced evolution of BabyAGI featuring enhanced task management prompt for comprehensive information analysis and synthesis. Built on GPT-4 architecture supporting multiple functions within single task management prompt. Introduces significantly more complex task management compared to original BabyAGI, providing foundation for sophisticated AI applications despite slower processing speeds.
BabyCatAGI
Lightweight 300-line modification of BabyBeeAGI with task creation agent running once, execution agent looping through tasks, and task dependencies for pulling relevant results. Features two tools (search and text completion), mini-agent functionality, and integrated search tool combining search, scrape, chunking, and extraction. Results synthesized into summary reports.
BabyDeerAGI
Compact 350-line BabyAGI modification optimized for GPT-3.5-turbo eliminating GPT-4 requirement. Features parallel task execution for faster performance, user input tool integration, query rewriting in web search tool, and automatic result saving capabilities.
BabyElfAGI
Expanded 895-line modification of BabyDeerAGI introducing Skills class for creating new capabilities, dynamic task list with vector search, beta reflection agent, and self-modifying code abilities allowing it to read, write, and review its own implementation.
BabyCommandAGI
CLI-LLM integration agent based on BabyAGI that executes shell commands for automatic programming and environment setup. Successfully demonstrated Flutter environment installation, app creation and launch on Linux containers. Handles general tasks including text generation, code writing, language translation. Requires virtual environment like Docker due to system modification risks. GPT-4 or higher recommended.
BabyFoxAGI
BabyElfAGI modification with self-improving FOXY method task lists that store final reflections and pull relevant insights to guide future runs. Features novel chat UI with parallel tasks panel, enabling multiple simultaneous task requests. New skills include DALLE image generation with prompt assist, Deezer music player, Airtable search integration, and Startup Analyst function. Chat can use single skills quickly or chain multiple skills via tasklist.
BambooAI
Data exploration and analysis tool for non-programmers operating in continuous loops. Mixes different models with varied capabilities, token costs, and context windows per task. Maintains conversation memory, builds dynamic prompts using Pinecone vector DB context, breaks down codified responses into step-by-step algorithms, generates and debugs Python code for analysis/modeling/plotting with auto-correction, ranks answers, requests user feedback, and builds knowledge-base from rankings.
BeeBot
Early-stage autonomous agent project under active development targeting wide range of task categories. Built by AutoPackAI for handling diverse productivity and general-purpose operations. BeeBot is currently a work in progress and should be treated as an early stage research project. Its focus is not on production usage at this time.
Blinky
Open-source AI debugging agent for VSCode inspired by SWE-agent. Uses LLMs to identify and fix backend code errors through VSCode API integration, Language Server Protocol (LSP), and print statement debugging to triangulate and address bugs in real-world systems.
Bloop
GPT-4 powered semantic code search engine for Rust and TypeScript codebases. Features precise code navigation built on stack graphs and scope queries, fast Rust-written regex matching engine, natural language and filtered query support for local and remote repositories, change staging capabilities, and downloadable app via GitHub.
BondAI
Highly capable autonomous AI agent with easy-to-use CLI, RESTful/WebSocket API, and pre-built Docker image. Supports all GPT-N, Embeddings and DALL-E OpenAI models plus Azure OpenAI Services. Features powerful code interpreter capabilities, Postgres DB data query integration, telephony support via BlandAI, stock trading via Alpaca Markets, and Gmail/Google Search integration. Simple pip install and commands to start CLI or server.
.gif)
bumpgen
AI agent for automated npm dependency management and upgrades. BUMPs package versions and GENerates code fixes for breaking changes. Supports gpt-4-turbo. Simple global npm install and easy command-line usage for dependency updates with automatic code adaptation.
Cal.ai
Open-source scheduling assistant built on Cal.com infrastructure. Books meetings, summarizes weekly schedules, and finds time with others using natural language. Responds flexibly to unseen tasks like 'move my second-last meeting to tomorrow morning'. Powered by GPT-4 and LangChain Agent Executor.
CAMEL
Communicative Agents for 'Mind' Exploration architecture - open-source library for building autonomous and communicative agent systems. Comprehensive framework for creating agents capable of complex communication patterns and collaborative behaviors.
ChatArena
Multi-Agent Language Game Environments for LLMs focused on developing AI communication and collaboration capabilities. Provides general framework for building interactive environments for multiple LLMs, collection of pre-built and community-created environments, and user-friendly interfaces with Web UI and command-line options.
ChatDev
Virtual software company powered by intelligent agents assuming roles like CEO, CPO, CTO, programmer, reviewer, tester, and art designer. Agents collaborate in organizational structure through functional seminars on design, coding, testing, and documentation. Provides accessible, modular, extensible LLM-based platform for studying collective intelligence with extensive customization for tailoring development processes, defining phases, and establishing roles. Committed to open-source principles.
ChemCrow
Open-source LangChain-based agent for accurate solution of reasoning-intensive chemical tasks. Integrates 13 expert-designed tools to augment LLM chemistry performance. Uses ReAct format (Thought, Action, Action Input, Observation). Human expert evaluations show ChemCrow significantly outperforms GPT-4 alone in chemical correctness and task completion, highlighting limitations of LLM self-evaluation in specialized domains.
Clippy
Development agent designed to build code for or with users. Capable of autonomous planning, writing, debugging, and testing for some projects. Optimal usage involves reviewing its work and providing iterative feedback for harder tasks.
CodeFuse-ChatBot
Intelligent assistant serving entire software development lifecycle powered by Multi-Agent Framework. Works with DevOps Toolkits, Code & Doc Repository RAG, and comprehensive development tools. Covers full spectrum from planning to deployment.
Cody by ajhous44
AI assistant for interactive codebase querying using natural language. Utilizes vector embeddings, intelligent chunking, and OpenAI language models to enable efficient and intuitive code navigation through semantic understanding.
Cody by Sourcegraph
AI code assistant that writes code and answers questions by reading your entire codebase and understanding the code graph. Provides intelligent code generation and contextual answers based on comprehensive codebase analysis.
Continue
Open-source autopilot for software development bringing ChatGPT capabilities to VS Code. Answers coding questions, enables natural language editing, and generates files from scratch. Streamlines development workflow with intelligent code assistance.
CrewAI
Framework for orchestrating role-playing autonomous AI agents that work together seamlessly, tackling complex tasks through collaborative intelligence. Built on LangChain, empowering engineers to harness collective AI agent power. Features modular design with agents, tools, tasks, processes, and crews. Each agent has specific roles, background stories, goals, and memories.
data-to-paper
Framework for systematically navigating AI through complete end-to-end scientific research, from raw data to comprehensive, transparent, human-verifiable research papers. Guides interacting LLM and rule-based agents through conventional scientific paths including data annotation, hypothesis creation, literature search, code writing and debugging, result interpretation, and step-by-step paper writing. Enhances transparency, traceability, and verifiability while allowing scientist oversight.
Databerry
No-code platform for creating AI chatbots trained on custom data. Pivoted to Chaindesk. Easy setup with model selection, data configuration, and deployment to website, Slack, Crisp, or Zapier. Built with Next.js, Joy UI, LangchainJS, PostgreSQL, Prisma, and Qdrant. Features streamlined customer support, team onboarding, data loading from multiple sources, user-friendly datastore management, and secured API endpoints.
DemoGPT
Leverages LLMs to provide fast and effective demo creation for applications of any purpose. Automates prototyping process, making it efficient and time-saving. Understands and processes prompts to generate relevant applications. Integrated with LangChain for code generation through iterative documentation parsing using Tree of Transformations approach. Roadmap includes constant improvements based on user feedback and addressing hallucination challenges.
DevGPT
AI team that builds microservices based on your ideas. Automated development team comprised of virtual Product Manager, Developer, and DevOps. Creates tailored microservices through advanced artificial intelligence, making software development seamless and efficient. Covers every aspect from concept to deployment.
Devika
Agentic AI Software Engineer that understands high-level human instructions, breaks them into steps, researches relevant information, and writes code to achieve given objectives. Competitive open-source alternative to Devin by Cognition AI.
Devon
Open-source alternative to Devin developed by Entropy research. Agentic AI software engineering tool focused on autonomous development capabilities.
DevOpsGPT
AI-Driven Software Development Automation Solution combining LLMs with DevOps tools to convert natural language requirements into working software. Improves development efficiency, shortens cycles, and reduces communication costs for higher-quality software delivery. Features automated code generation and validation. Enterprise Edition includes existing project analysis, professional model selection with private deployment support, and integration with more DevOps platforms.
dotagent
Agent management system facilitating creation of robust AI applications and experimental autonomous agents through rich developer tools. Enables deployment across cloud, PCs, or mobile devices with Python or plain English integrations. Features powerful prompt compiler for advanced Language Model control, optimizing response generation. Allows seamless export of agents into portable files for execution in any environment, with optional Agentbox for optimized computing resource management in sandboxed environment.
Eidolon
Open-source SDK for building AI agents with pluggable and modular component architecture. Enables rapid development and deployment of multi-agent systems through flexible, customizable framework.
English Compiler
AI-based compiler converting English-based markdown specifications into functional code. Based on the premise that great projects start with detailed functional specifications typically written in natural language. Compiles written specifications directly to code, envisioning a future where software development is driven primarily by written text.
evo.ninja
AI agent that adapts itself in real-time based on tasks at hand. Utilizes pre-defined agent personas tailored to specific task domains. Each iteration of execution loop selects and adopts the persona fitting the task best, making it uniquely adaptive.
FastAgency
Open-source framework designed to accelerate transition from prototype to production for multi-agent AI workflows. For AutoGen framework users, enables seamless scaling of Jupyter notebook prototypes into fully functional production-ready applications. Features multi-framework support, unified programming interface, and powerful API integration capabilities, streamlining deployment while maintaining flexibility and performance.
Flowise
Open-source low-code tool for developers to build customized LLM orchestration flows and AI agents. Enables rapid development of sophisticated AI applications through visual interface and modular components.
Friday
AI developer assistant capable of creating complete Node.js projects with unlimited prompts. Provides core prompt for application foundation with ability to add unlimited sections representing specific app parts. Utilizes GPT-4 for assistance, optimized with GPT-4-32k for improved speed and results. Requires 2 small requests for app base and 1 request per section. Employs esbuild behind the scenes for every created app.
GeniA
Engineering platform AI team member capable of working alongside you in production environment, executing tasks in dev and cloud environments including AWS, Kubernetes, Argo, GitHub, etc. Allows enhancement through integration of custom tools and APIs. Features Slack App Bot integration and supports GPT-3.5 and GPT-4.
Godmode
Web platform inspired by AutoGPT and BabyAGI for conducting various tasks through nice UI. Capabilities include ordering coffee at Starbucks, performing market analysis, and finding and negotiating leases. Supports GPT-3.5 and GPT-4.
GPT Discord
Robust all-in-one GPT interface for Discord. Supports multi-modality image understanding, code interpretation, advanced data analysis, Q&A on custom documents, internet-connected chat with Wolfram Alpha and Google access, AI-moderation, DALL-E image generation, and much more. Features code execution and environment manipulation by E2B. Supports OpenAI models.
GPT Engineer
AI agent that generates entire codebases based on prompts using GPT-4. Asks for clarification before constructing complete code. Easy to adapt, extend, and customize agent behavior by editing identity files. Stores communication history for each step in logs folder, allowing message log reruns.
GPT Migrate
Easily migrates codebases from one framework or language to another. Pick from different LLMs, generate and run unit tests for new codebase, select source and target languages, and customize agent workflow (setup, migrate, test). Team working on adding benchmarks for agent performance measurement.
GPT Pilot
AI agent that codes entire apps from scratch while developers oversee implementation. Dev tool writing scalable apps as you supervise code being written. Research project exploring how GPT-4 can generate fully working, production-ready apps. Main idea: AI can write most code for an app (maybe 95%) but the rest 5% requires a developer until full AGI.
GPT Researcher
GPT-based autonomous agent performing online comprehensive research on any topic. Produces detailed, factual, and unbiased research reports. Offers customization for focusing on relevant resources, outlines, and lessons. Addresses speed and determinism issues through parallelized agent work. Inspired by AutoGPT and Plan-and-Solve paper. Uses planner and execution agents - planner generates research questions, execution agents find related information.
GPT Runner
Conversation with selected files without embedding or vector database. AI Prompt Storybook for managing AI presets with teams. Supports any IDE and language developer. Provides CLI for web and VSCode extension, Jetbrains plugin coming soon. Privacy-first with all data local. Supports OpenAI and Anthropic (Claude-2). Multi-language support.
GPTSwarm
Graph-based framework for LLM-based agents providing two high-level features: building agents from graphs and enabling customized automatic self-organization of agent swarms with self-improvement capabilities. Unifies diverse prompt engineering techniques by describing LLM agents as computational graphs. Each node implements a function to process multimodal data or query LLMs. Each edge describes information flow. Features automatic graph optimizers for node-level prompt refinement and agent orchestration improvement through graph connectivity changes.
IX
LangChain-based platform for building, debugging, and deploying collaborative agents and cognitive workflows. Includes tools to build and deploy agent fleets that collaborate to complete tasks. Both editor and runtime - no-code graph style editor for designing agents, chains, tools, retrieval functions, and collaborative workflows. Features horizontally scaling agent worker fleet, multi-user multi-agent chat interface, smart input with @mentions and file references, Chroma and other vector database support, and supports OpenAI API, Anthropic, PaLM, and LLama models. Powered by LangChain with easily extended component library.
JARVIS
System connecting LLMs with ML community. Task Planning: ChatGPT analyzes user requests to understand intention and disassemble into solvable tasks. Model Selection: ChatGPT selects expert models from Hugging Face based on descriptions to solve planned tasks. Task Execution: Invokes and executes selected models, returning results to ChatGPT. Response Generation: ChatGPT integrates all model predictions and generates responses.
Langroid
Intuitive, lightweight, extensible and principled Python framework to easily build LLM-powered applications. Set up Agents, equip them with optional components (LLM, vector-store, methods), assign tasks, and have them collaboratively solve problems through message exchange. Multi-Agent paradigm inspired by Actor Framework. Fresh take on LLM app-development with simplified developer experience - does not use Langchain. Works with most commercial/remote and open/local LLMs. Supports multi-agent, multi-LLM systems, OpenAI function-calling and native ToolMessage. Batteries-included with vector-databases for RAG, caching, logging. Specialized agents available: DocChatAgent, SQLChatAgent, TableChatAgent.
Lemon Agent
Plan-Validate-Solve (PVS) Agent for accurate, reliable and reproducible workflow automation. Standalone supervised Plan and Solve Agent specialized in read and write operations on tools like GitHub, HubSpot, or Airtable. Separation of tasks with human-in-the-loop interactions: holds Planner and Solver Agents for focus and accuracy. Asks for approval at relevant workflow steps. Unlimited configuration options for workflow definition including permission prompts and custom actions. UI flexibility with any interface or built-in CLI. Model and framework agnostic operations.
LLM Agents
Minimalistic library for building agents leveraging large language models to automate tasks through command loops and tool integrations. Executes Python code in REPL environment, conducts Google and Hacker News searches. Iterates through Thought, Action, Observation, New Thought cycle based on tool output. Dynamically appends information to prompts for informed agent decision-making.
LLM Stack
No-code platform to build LLM Agents, workflows and applications with your data. Supports all major model providers including OpenAI, Cohere, Stability AI, and Hugging Face. Build generative AI agents like AI SDRs, Research Analysts, RPA Automations without code. Connect agents to internal/external tools, search web, or browse internet.
Local GPT
Chat with documents without compromising privacy. LocalGPT is an open-source initiative that allows you to converse with your documents without compromising your privacy. Inspired by privateGPT, allows using your own documents as an information source. Chat with your documents on your local device using GPT models. No data leaves your device and 100% private. With everything running locally, you can be assured that no data ever leaves your computer. Dive into the world of secure, local document interactions with LocalGPT. Model: Vicuna-7B. Using InstructorEmbeddings. Both Embeddings as well as LLM will run on GPU. It also has CPU support if you do not have a GPU. Built with Langchain.
Loop GPT
Modular and extensible Python package reimplementing Auto-GPT framework. Supports GPT-3.5-turbo and GPT-4 with plug-and-play API. Features easy integration of new capabilities from Python code without config files, minimal prompt overhead for token efficiency, human-in-the-loop course correction, and full state serialization enabling agent state saves including memory and tool states without external databases.
L2MAC
Multi-agent generation framework creating extensive unbounded outputs like entire codebases or books from single input prompts. Achieves state-of-the-art generation for large codebase tasks ranking top 3 in HumanEval coding benchmark. Detects invalid code and failing unit tests with automatic error correction. Maintains complete file-store memory enabling LLM agents to read/write files. Follows exact prompt programs bypassing fixed context token limits through iterative generation. Published at ICLR 2024.
Maige
Codebase agent running automatically on new issues and pull requests. Core features include labelling, assigning, and answering questions. Can search entire codebase, spin up sandbox to run scripts, and write basic code. Natural-language workflow automation for GitHub repositories.
Magick
Full-suite model-agnostic AIDE (AI Development Environment) for creating, deploying, scaling, and monetizing AI agents and prompt chaining. Enables building systems like BabyAGI within an hour. Features real-time graph execution visualization, thought process observation, and flow understanding. Built specifically for rapid cognitive architecture development and scalable event-driven autonomous agents. Visual development of autonomous agents.
MemFree
Open-source hybrid AI search engine delivering instant accurate answers from Internet, bookmarks, notes, and docs. Features one-click Chrome bookmarks sync and indexing, multiple traditional search engine source support, self-hosted super fast serverless vector database, self-hosted local embedding and rerank service, full open-source code, and one-click production deployment.
MemGPT
System intelligently managing different memory tiers in LLMs to effectively provide extended context within limited context window. Features chat with local files or SQL database, and creates perpetual chatbots with self-editing memory capabilities. Hierarchical memory architecture for enhanced context management.
Mentat
Command-line AI coding assistant that coordinates edits across multiple locations and files. Unlike Copilot, handles multi-file edits. Unlike ChatGPT, already has full project context without copy-pasting. Direct command-line integration for coding tasks.
MetaGPT
Multi-agent framework that from one-line requirement returns PRD, Design, Tasks, or complete Repo. Assigns different roles to GPTs forming collaborative software entity for complex tasks. Outputs user stories, competitive analysis, requirements, data structures, APIs, documents. Internally includes product managers, architects, project managers, engineers. Provides entire software company process with carefully orchestrated SOPs. Core philosophy: Code = SOP(Team). Incorporates efficient human workflows as meta programming into LLM multi-agent collaboration.
Mini AGI
Minimal general-purpose autonomous agent based on GPT-3.5-Turbo/GPT-4. Can analyze stock prices, perform network security tests, create art, and order pizza. Combines robust prompt with minimal toolset, chain-of-thoughts, and short-term memory with summarization. Features inner monologue and self-criticism capabilities.
Multiagent Debate
Implementation of paper 'Improving Factuality and Reasoning in Language Models through Multiagent Debate'. Treats different instances of same language models as 'multiagent society' where individual instances generate and critique each other's outputs. Final answers after this procedure show improved factual accuracy and better reasoning question solving. Demonstrates quantitative improvements in multiple domains of reasoning and factual validity.
Multi GPT
Experimental open-source attempt at fully autonomous GPT-4 where multiple expertGPTs collaborate on tasks. Each expert has own short and long-term memory with inter-agent communication abilities. Features include task setting with expert collaboration, internet access for searches, memory management, GPT-4 instances for text generation, access to popular platforms, and file storage/summarization with GPT-3.5.
MutahunterAI
Generates unit tests specifically targeting code vulnerabilities using mutation testing engine, unlike copilots that blindly generate tests. Boosts developer productivity by targeting exact code weaknesses. Supports all major languages, local or CI/CD integration, and works with your own LLM APIs for privacy. Open-source AI for accelerating developer productivity and code security.
NLSOM
Natural Language-Based Societies of Mind featuring societies and communities of agents including LLMs, NN-based experts, APIs, and role-players all communicating in natural language. Agents solve tasks through collaborative 'Mindstorm' process involving mutual interviews. Modular architecture for easy component addition. Based on Marvin Minsky philosophy: intelligence stems from vast diversity, not single perfect principle.
OpenAgents
Open Platform for Language Agents in the Wild serving as ChatGPT Plus replica for researchers, developers, and general users. User-centric with Chat Web UI, productive agents, and online demo. Fully open-sourced full-stack solution with easy deployment. Extensible supporting multiple LLMs, tools, and agent methods.
OpenAGI
Open-source AGI R&D platform enabling agents for benchmark and open-ended tasks. Powered by GPT-4, Vicuna, LLaMA, Flan-T5. Supports multi-modality (text, image, video, audio) tool learning and task solving. Features linear and non-linear task decomposition, benchmark and open-ended task solving, easy evaluation protocols, and RLTF (Reinforcement Learning from Task Feedback) for continuous self-improvement.
OpenDevin
Open-source project born from desire to replicate, enhance, and innovate beyond original Devin model. Engages open-source community to tackle Code LLM challenges in practical scenarios, producing works that significantly contribute to community and pave way for future advancements. Motto: 'Code Less, Make More'.
Open Interpreter
Open-source interpreter letting LLMs run code on your computer to complete tasks. Runs locally enabling tasks like PDF summarization, dataset visualization, and browser control. Works through ChatGPT-like terminal interface.
Pezzo
Development toolkit streamlining prompt design, version management, publishing, collaboration, troubleshooting, and observability. For technical and non-technical users. Features centralized prompt management, streamlined design/publishing/versioning, detailed execution history with stats (duration, costs), retroactive troubleshooting with time travel, cost transparency, and 90% code overhead reduction via Pezzo Client.
Private GPT
Tool for completely private document interaction without internet connection. Built with LangChain, GPT4All, LlamaCpp, Chroma, and SentenceTransformers. Test project validating feasibility of fully private question-answering solution using LLMs and vector embeddings. Not production-ready.
PromethAI
Personalized AI assistant decomposing problems and offering solutions with Agent actions for workflow automation. Helps users reach solutions through categorical request decomposition, features dynamic UX/UI avoiding prompting, supports voice input, provides query results with action automation, and remembers preferences for choice optimization. Powered by Langchain with decomposable async prompts, vector DB, Redis cache. Built with Flutter/Dart, connected to Zapier NLP.
React Agent
Experimental autonomous agent using GPT-4 to generate and compose React components from user stories. Stack includes React, TailwindCSS, TypeScript, Radix UI, Shandcn UI, and OpenAI API. Takes user story text and generates multiple React components for relevant screens based on atomic design principles.
Self-operating computer
Enables multimodal models to operate computers using same inputs/outputs as human operators. Model views screen and decides series of mouse and keyboard actions to reach objectives. Autonomous computer control through visual understanding and action planning.
Smol developer
Your own junior developer deployed in seconds via E2B. Human-centric coherent whole program synthesis. Features 200 LOC (half English), 100k context for summarizing content and codebases, markdown as prompting DSL, copy-paste errors/curl output as prompts, CSS animation by description. GPT-4 significantly better than GPT-3.5/Claude for codegen. Allows develop, debug, and decompile.
Stackwise
VS Code extension that automatically writes and imports Node.js functions eliminating context switching. Open-source function collection where you explain what function should do and AI builds it. No more documentation hunting or ChatGPT back-and-forth. Pure functionality within your code.
Superagent
Tool for creating agents without coding, simplifying LLM Agent configuration and production deployment. Built with zero third-party dependencies to proprietary tech. Features built-in memory, document retrieval via vector DBs, powerful tools, webhooks, cron jobs. Supports action agents and plan-and-execute agents.
SuperAGI
Open-source autonomous AI framework enabling development and deployment of autonomous agents. Features agent provisioning/spawning/deployment, tool-based capability extension, concurrent agent execution, GUI, action console, multiple vector DBs, multi-modal agents, trajectory fine-tuning, performance telemetry, optimized token usage, memory storage, looping detection heuristics, and resource management. Open source framework with closed-source infrastructure.
Suspicion Agent
Research on playing imperfect information games with Theory of Mind aware GPT-4. Paper delves into GPT-4's learned knowledge applicability for imperfect information games. Demonstrates agent capabilities in games requiring hidden information reasoning.
SWE Agent
Open-source Devin alternative achieving 12.3% on FULL SWE-bench (vs Devin's 13.84% on 25% of test set). Works through specialized terminal allowing file opening/scrolling/searching, line editing with syntax check, and test writing/execution. Custom-built interface critical for performance. Uses Agent-Computer Interface (ACI) constraining commands to run/look/edit code and submit to GitHub. Code undergoes linter checks with feedback loop. Reads 100 lines at a time for better LM comprehension. Released by Princeton NLP team.
Sweep
GitHub assistant helping fix small bugs and implement small features. Install via button, add repository, create ticket (e.g. writing tests), prepend with 'Sweep:' and let Sweep handle the rest. Automated code generation and pull request creation.
Taxy AI
Full browser automation using GPT-4 to control browser and perform repetitive actions. Allows ad-hoc instruction definition. Future support for saved and scheduled workflows. Early stage with waitlist access.
Teenage AGI
BabyAGI-inspired agent with infinite memory recall that thinks before acting and persists memory after shutdown. Uses GPT-4, OpenAI, and Pinecone for memory persistence. Process: vectorizes queries, searches relevant memories, thinks about actions, stores thoughts, generates outputs, and stores query-answer pairs. Memory lives on Pinecone with index counter for persistence.
UFO
UI-Focused dual-agent framework by Microsoft fulfilling user requests on Windows OS by seamlessly navigating and operating within individual or spanning multiple applications. Native Windows application control and automation.
Vanna.AI
Open-source Python-based AI SQL agent trained on your schema writing complex SQL in seconds. AI-driven BI assistant using LLMs via Retrieval-Augmented Generation. Two-step process: train RAG model on data, then ask questions returning SQL queries auto-running on database. Accuracy tied to training data volume. Database contents never sent to LLM - only schemas, documentation, queries. Model continuously improves with usage through training data augmentation.
Voyager
LLM-powered embodied lifelong learning agent in Minecraft continuously exploring world, acquiring diverse skills, and making novel discoveries without human intervention. Three key components: automatic curriculum maximizing exploration, ever-growing skill library of executable code for complex behaviors, and iterative prompting mechanism incorporating environment feedback, execution errors, and self-verification. Interacts with GPT-4 via blackbox queries bypassing fine-tuning need.
Web3 GPT
AI-powered platform for writing and deploying smart contracts directly to EVM-compatible blockchains. Enables users to connect their wallets for manual contract deployment, supports account abstraction for Web2/3 users without dedicated web3 wallets using a gas master account, and leverages ChatGPT to interact with and control Web3-GPT functionalities.
Westworld simulation
Multi-agent simulation library designed to simulate and optimize systems with multiple interacting agents. Features easy creation of grid and non-grid environments, custom objects through subclassing, spawner for random object generation, rigid body physics, agent behaviors including pathfinding, wandering, and vision range, automatic maze generation, layer integration to convert images to obstacles, and visualization with replay and export capabilities.
WorkGPT
GPT agent framework for invoking and integrating APIs, similar to AutoGPT and LangChain. Accepts a directive and array of APIs, then converses with AI until the directive is complete. Supports research tasks, web crawling, service ordering, and any API representable with OpenAPI files. Features OpenAI's function invocation for faster, more cost-effective, and reliable API chaining.
Wren
AI-powered data assistant providing natural language interface to databases, retrieving results and insights without SQL coding. Built on four principles: explainability with accurate SQL generation, interoperability across multiple data sources with standard interface, interactive dialogue for query clarification and refinement, and continuous learning from user interactions and feedback.
XAgent
Open-source experimental LLM-driven autonomous agent that automatically solves various tasks. Features emergence and autonomy transcending biases, safe execution in docker environments, expert-knowledge-free operation, user-friendly GUI and command line interfaces. Implements dual-loop mechanism with outer-loop for plan management and inner-loop for ReACT-based execution. Includes integrated tools: file editor, Python notebook, web browser, and shell.
yAgents
Agent-Builder Agent capable of designing, coding, and debugging its own tools. Built on LangChain framework for building, prototyping, and deploying AI-powered tools and agents easily. Features interactive command line interface for real-time feedback, session persistent memory for data preservation across sessions, and quick installation via pip. Designed for users of any technical background.
Yourgoal
Swift port of BabyAGI, an AI-powered task management system using OpenAI and Pinecone APIs to create, prioritize, and execute tasks. Creates tasks based on previous task results and predefined objectives, enabling autonomous goal-directed behavior in Swift applications.