A curated index of AI agent frameworks, security tools, and autonomous workflows for 2026. Covers offensive security, OSINT, coding assistants, and emerging protocols — with honest technical assessments.
Tip
➕ Know a tool that should be here? Add it via PR or open an issue.
To ensure clarity across all repository tables, we classify tools using the following parameters:
- Engine Classification:
- Local Inference: Requires access to local hardware resources (GPUs/TPUs) or is optimized for edge/local deployment using quantized models.
- API-based: Execution relies heavily or entirely on managed external endpoints (e.g., OpenAI, Anthropic, Gemini APIs).
- Hybrid: Can operate utilizing both local system intelligence and fallback/routing to external provider APIs.
- Deployment Mode: The primary interface for utilizing the tool (e.g., CLI, Web UI, Docker, Library / SDK).
- AI Agent Frameworks
- Programming and Coding Agents
- Agentic Workflows and Automation
- Ethical Hacking and Bug Hunting
- OSINT (Open Source Intelligence)
- PC Control and Desktop Automation
- Uncensored Models and Jailbreak
- Protocols and Standards
- Observability and Evaluation
- Local and Self-Hosted AI Agents
- Advanced Guides, Tips, and Prompts
- Enterprise and CI/CD Agents
- Agent Directories and Marketplaces
- Specialized Agents
- Memory and Persistence for Agents
- Evaluation Benchmarks
- Simulations of Societies and Virtual Worlds
- Security, Guardrails, and Alignment
- Edge and Local AI Agents (Mobile/IoT)
- Web3 and Crypto Agents
- Voice and Real-Time Audio Agents
- Data Engineering and Analytics Agents
Base infrastructure to build autonomous agents.
| Repository | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| LangChain | Graph-based routing for NLP operations. Executes abstraction wrappers over proprietary REST APIs and local inferences. | Python/TS | Hybrid | Library | Link |
| LangGraph | Stateful multi-actor orchestrator mapping iterative nodes via Directed Cyclic Graphs (DCGs) with check-pointing. | Python/TS | Hybrid | Library | Link |
| LightAgent | Lightweight Python agent framework with tool calling, memory, MCP/SSE integration, deterministic workflows, and LightSwarm collaboration. | Python | Hybrid | Library | Link |
| CrewAI | Role-delegated DAG executor orchestrating task arrays through inter-agent message passing and memory state sharing. | Python | Hybrid | Library | Link |
| AutoGen | Multi-agent converse framework utilizing socket-based RPC loops for distributed execution safely sandboxed in Docker. | Python | Hybrid | Web UI | Link |
| MetaGPT | Waterfall SOP emulator mapping standard operational procedures to agent hierarchies for complex codebase iteration. | Python | Hybrid | CLI | Link |
| AutoGPT | Recursive heuristic loop agent generating execution plans with local filesystem caching and vector embeddings. | Python/Docker | Hybrid | Web UI | Link |
| Dify | Backend-as-a-Service architecture unifying API gateways for LLM requests with embedded vector DB routing. | Docker/Python | Hybrid | Web UI | Link |
| Bifrost | Open-source Go gateway exposing an OpenAI-compatible API for multi-provider routing, fallbacks, MCP, and request observability. | Go | Hybrid | Docker | Link |
| Camel-AI | Sociodynamic role-playing simulator executing prompt-based adversarial networks through isolated inference channels. | Python | Hybrid | Library | Link |
| SmolAgents | Minimal dependency wrapper supporting local safetensors parsing and direct execution of inline python tools. | Python | Hybrid | Library | Link |
| SuperAGI | Dockerized scalable infrastructure supporting distributed concurrent agent workers and shared vector store backends. | Docker/Python | Hybrid | Web UI | Link |
| SandBase Harness | Self-hosted runtime for AI coding agents with isolated workspaces, policy-gated tool execution, MCP support, and auditable task receipts. | TypeScript/Docker | Hybrid | Docker | Link |
Development, review, and scaling of source code managed by AI.
| Repository | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| Aider | Command-line AST manipulator supporting real-time git-diff integration and context-aware repository embeddings. | Python | Hybrid | CLI | Link |
| OpenHands | Container-isolated execution environment with bidirectional headless browser manipulation and pseudo-TTY interfaces. | Docker/Python | Hybrid | Web UI | Link |
| Void | Electron-based IDE fork with embedded inference routing, preventing telemetry leakage via local LLM socket binding. | TypeScript | Local Inference | IDE Extension | Link |
| Continue | IDE proxy layer routing context payloads and FIM requests to strictly configured local/remote inference endpoints. | TypeScript | Hybrid | IDE Extension | Link |
| SWE-agent | Automated patching daemon mapping GitHub issues to local repository diffs utilizing isolated bash evaluation. | Python | Hybrid | CLI | Link |
| GPT-Engineer | Scaffolding compiler converting high-level markdown specs to fully initialized workspaces via sequential AST generation. | Python | API-based | CLI | Link |
| Agentlas OS | Local-first agent OS for portable agent teams, cross-host orchestration, MCP/A2A, and verification gates. | Python | Hybrid | CLI | Link |
| Cline | IDE-embedded background worker interacting with the Language Server Protocol and filesystem watchdogs for code mutating. | TypeScript | Hybrid | IDE Extension | Link |
| Mentat | CLI abstraction operating over raw file streams and repository graphs for cross-file variable reference propagation. | Python | Hybrid | CLI | Link |
| molt | Coding agent that won't say done on a false claim; verification on disk with receipts for accepts and refusals. Terminal and desktop. | TypeScript | Hybrid | CLI | Link |
| Plandex | Go-based background orchestrator executing iterative refactoring tasks through detached daemon processes and git tracking. | Go | Hybrid | CLI | Link |
| Tabby | Rust-optimized inference engine providing low-latency FIM (Fill-In-the-Middle) payload processing for code completion. | Rust/Docker | Local Inference | Docker | Link |
| fractal | Hierarchical coding-agent runtime executing recursive loops in per-node Git worktrees with persistent SQLite state and configurable iteration, depth, child, cost, and time limits. | Python | API-based | CLI | Link |
| Atomic Agent | Coding and agentic runtime running open-weight models entirely on the local machine through a llama.cpp fork, with no account or API key required. Ships 56 built-in tools (browser, filesystem, git, memory, vision), MCP support, and a five-layer local memory system. Currently a developer preview: APIs, commands, and config are still moving. | TypeScript | Local Inference | CLI | Link |
| YYLO | Command-line orchestrator for coding agents, repeatable workflows, and receipt-backed repository changes. Each task freezes the protected target SHA and runs in a dedicated branch/worktree behind typed task, validation, merge, and release-readiness boundaries; the merge queue owns risk-based review. Installs via npm as @yylo/cli and orchestrates Pi and Codex subagents. | Python/TS | API-based | CLI | Link |
| CodePilot | Embeddable Python runtime for autonomous coding agent with ReAct execution, filesystem, terminal, MCP and Python tools. Uses text based code-as-interface protocol rather than native JSON tool calling. | Python | Hybrid | Library | Link |
| Sillage | Self-hosted mobile-first web UI that drives the native Claude Code and Codex CLIs on your own machine, with server-side session journals that outlive the client, full-text search across conversations, an IDE panel (files, editor, diffs, terminal), and a board the agents read through its own MCP server. Installable PWA with push. | TypeScript | Hybrid | Web UI / Docker | Link |
| Zaru | Terminal agent that runs model-requested tool calls behind per-call permission prompts, re-runs project-declared validator commands until they pass, and logs every step to a JSONL transcript. Pre-alpha, Linux only. | Rust | Hybrid | CLI | Link |
Integration of agents with APIs, nodes, and conditional process flows.
| Repository | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| API.market MCP Gateway | Proprietary hosted gateway with five tools for API discovery, execution, usage and subscriptions; OAuth or API keys; per-API free tiers and paid plans. | TypeScript/MCP | API-based | Web UI | Gateway |
| Aident Loadout | Proprietary hosted MCP server (Streamable HTTP, OAuth sign-in) and CLI that let MCP clients search and run actions in connected third-party apps and use 400+ prebuilt Skills; stored API keys stay in Aident Vault and each action is logged in Audit; account required, pay-as-you-go. | TypeScript/MCP | API-based | CLI / Web UI | Link · Docs |
| GitHub Agentic Workflows | CI/CD pipeline wrapper converting natural language to GitHub Action YAML files triggering matrix runners. | YAML/TS | API-based | Docker | Link |
| Composio | Authentication middleware providing OAuth2 bridging and strictly-typed endpoint mapping between LLM and 3rd party APIs. | Python/TS | Hybrid | Web UI | Link |
| n8n | Node-based visual workflow engine running native JS sandbox execution and asynchronous event-driven RPC triggers. | TS/Docker | Hybrid | Web UI | Link |
| PydanticAI | Type-safe middleware guaranteeing structured JSON output adherence through hardcoded schema validation and retry loops. | Python | Hybrid | Library | Link |
| Agent-Orchestrator | Automated PR lifecycle controller validating diffs against test suites using headless execution inside ephemeral CI. | Python | API-based | CLI | Link |
| LLM Workflow Engine | Pipeline executor processing sequential prompts and branching logic through deterministic finite automata state machines. | Python | Hybrid | Web UI | Link |
| Bee Agent Framework | Enterprise-grade SDK enforcing hardened guardrails, strict data plane routing, and audit-logging compliance. | TypeScript | Hybrid | Library | Link |
| Awesome LLM Apps | Technical boilerplate and infrastructure definitions for production RAG microservices and scalable vector integrations. | Markdown | N/A | Document | Link |
| Agent Coordinator | Per-user Codex skill that runs bounded dependency-graph work inline or through optional specialists, stores revisioned local state, rejects concurrent live work with overlapping declared write scopes, reconciles uncertain operations before retry, and reruns authored closeout checks. | Python/Codex Skill | API-based | CLI | Link |
Penetration testing, vulnerability analysis, and offensive routines.
| Repository | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| HexStrike AI | Execution of native offensive dependencies. Requires root privileges for raw sockets and offers basic EDR evasion. | Python | API-based | CLI | Link |
| Claude Bug Bounty | AST repository analyzer. Generates dry structured payloads (does not auto-execute dynamic payloads). | Python/TS | API-based | CLI | Link |
| PentestAgent | Black-Box operator using network sub-processes. Unencrypted traffic detectable by standard IDS; statically auto-generated payloads. | Python | Hybrid | CLI | Link |
| Darkmoon | Autonomous offensive engine with per-technology sub-agents (web, cloud, AD, K8s); validates findings via real exploit execution; Privacy Gateway masks sensitive data from the model. | Python | Hybrid | CLI | Link |
| Pentagi | Kill Chain execution delegated to Docker. Lacks obfuscation or native L7 firewall bypass. | Docker/Go | Hybrid | Web UI | Link |
| CAI | TCP/UDP services inference. Passive protocols without offensive exploitation payload generation. | Python | Hybrid | Library | Link |
| Awesome AI Cybersecurity | Structured references for Red Team modeling. Purely documentary passive environment. | Markdown | N/A | Document | Link |
Extraction, filtering, and summarization of information from public sources.
| Repository | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| OSINT AI Agent | Continuous domain reconnaissance sequences. Executes massive queries to APIs (Shodan/Censys) susceptible to cap limits. | Python | API-based | CLI | Link |
| AI OSINT | Automated tracking of Google Dorks syntax. Extracts extensive indexes; requires proxy rotation to evade bans. | Python | API-based | CLI | Link |
| AI-Resources | Scripts for static headless scraping (Selenium). Extracts profile networks but struggles with dynamic JS validations and CAPTCHAs. | Python | Hybrid | CLI | Link |
| Awesome Cybersecurity Agentic AI | Compilation of repositories and passive analysis schemas for OSINT sources. | Markdown | N/A | Document | Link |
| Ubikron Graph Architecture | Correlation of intelligence entities via graph databases. Maps iterative B2B relationships. | Graph/Neo4j | Local Inference | Web UI | Link |
| Awesome OSINT Digester | Comprehensive directory of OSINT sources for manual intelligence gathering and agentic ingestion. | Python | Hybrid | CLI | Link |
Multimodal interactors for graphical interfaces via mouse and keyboard clicks.
| Repository | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| Self-Operating Computer | Mouse and keyboard management through multimodal models. | Python | API-based | CLI | Link |
| Agent S | Asynchronous progressive processing for complex OS tasks. | Python | Hybrid | CLI | Link |
| Browser-use | Direct interface between LLMs and web browser instances. | Python | API-based | CLI | Link |
| Agent QA | Natural-language web, Android, and iOS UI testing with execution memory; source-available under FSL-1.1-ALv2. | TypeScript | Hybrid | CLI / Web UI | Link |
| CUA | Desktop container simulation environment. | Docker/Py | Local Inference | Docker | Link |
| Bytebot | Control of containers with natural language commands. | Python | Hybrid | Docker | Link |
| Agent-Desktop | Structured native adapter for GUI applications. | Rust | Local Inference | CLI | Link |
Red Teaming, filter evasion, and tensor alteration in base models.
| Repository | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| Heretic | Orthogonal tensor ablation. VRAM: 12GB+ (8B), 24GB+ (32B). Formats: Safetensors. Framework: HF Transformers/Native PyTorch. | Python | Local Inference | CLI | Link |
| USB-Uncensored-LLM | Encapsulated binary environment. VRAM: 8GB to 16GB. Supported formats: GGUF (Q4_K_M, Q5_K_M). Framework: llama.cpp. | Python/CPP | Local Inference | Docker | Link |
| Dolphin Mixtral | Derived MoE model weights. VRAM: 32GB+ (GGUF Q4) or 2x24GB (AWQ/GPTQ). Frameworks: vLLM (for AWQ), Ollama (GGUF). | Weights | Local Inference | Web UI | Link |
| Awesome Jailbreak on LLMs | Iterative attack vectors (GCG) and heuristic injection over model context windows. | Markdown | N/A | Document | Link |
| LLM-Jailbreaks | Logical exploits based on overflow or semantic validators and pre-prompts manipulation. | Text | Hybrid | Document | Link |
| Red Team AI Benchmark | Automated pipelines for robustness evaluation and filtering across MLaaS endpoints and local nodes. | Python | Hybrid | CLI | Link |
| Geneva | Native TCP/IP root manipulation to evade Layer 7 firewalls and Deep Packet Inspection (DPI) over LLM traffic. | Python | Local Inference | Docker | Link |
Interoperability protocols, tool-use specifications, and agent communication standards.
| Repository | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| MCP (Model Context Protocol) | Open standard by Anthropic for connecting AI models to external tools, data sources, and services via a JSON-RPC protocol. | TypeScript/Python | Hybrid | Library | Link |
| A2A (Agent-to-Agent) | Google's open protocol enabling inter-agent communication and task delegation across heterogeneous agent frameworks. | Python | Hybrid | Library | Link |
| Agent Protocol | Unified API standard by AI Engineer Foundation defining a consistent REST interface for agent creation, execution, and task management. | Python/TS | N/A | Library | Link |
| OpenAI Realtime API | WebSocket-based specification for low-latency streaming audio and multimodal agent interactions. | Python/TS | API-based | API | Link |
| LangChain Tool Standard | Standardized tool interface allowing cross-framework tool sharing between LangChain, CrewAI, AutoGen, and other agent runtimes. | Python/TS | Hybrid | Library | Link |
Tracing, monitoring, and debugging agent executions in production.
| Repository | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| Langfuse | Open-source LLM observability platform tracing agent runs, prompt versions, and cost metrics with self-hosted option. | TypeScript/Python | API-based | Web UI | Link |
| LangSmith | LangChain's native tracing and evaluation platform for debugging multi-step agent pipelines and regression testing. | Python/TS | API-based | Web UI | Link |
| Arize Phoenix | Open-source AI observability engine for tracing LLM calls, evaluating RAG quality, and detecting hallucinations. | Python | Local Inference | Jupyter | Link |
| Helicone | Open-source proxy layer logging LLM requests with caching, rate limiting, and cost analytics dashboards. | TypeScript | API-based | Web UI | Link |
| Braintrust | Evaluation framework for scoring agent outputs against golden datasets with automatic regression detection. | TypeScript | API-based | Web UI | Link |
| antiburn | Local desktop app that checks coding agent sessions for common causes of token burn: sessions too deep, overpowered subagents, broken caching, and unused MCP servers, skills, and tools. Supports Claude Code, Codex, Cursor, Copilot, Pi, and more. | Rust/Tauri | N/A | App | Link |
| OrcaReplay | Local record-and-replay proxy for agent runs: captures provider traffic (prompts, tool calls, responses) into a trace and replays the run from it without calling the provider again, for deterministic debugging and regression diffs. | TypeScript/Node | Hybrid | CLI | Link |
| Self-Auditing Agent | Public audit log for an agent's own claims: each recorded conclusion links a runnable command a third party can execute, so the process is reviewable rather than just its output. Ships gatecheck, a mutation tester that runs 175 mutations against the project's own validator and reports the ones it fails to catch. | Python | N/A | CLI | Link |
Privacy-first agents running entirely on local hardware without cloud API dependencies.
| Repository | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| Ollama | Lightweight daemon serving quantized LLMs locally with a REST API and model management for agent backends. | Go/C++ | Local Inference | CLI | Link |
| Jan | Desktop application running local LLMs with an embedded agent runtime and GPU-accelerated inference. | TypeScript/C++ | Local Inference | App/Web | Link |
| LocalAI | Drop-in OpenAI-compatible API server running on CPU/GPU with model quantization and function calling support. | Go/Python | Local Inference | Docker | Link |
| LM Studio | Desktop GUI for discovering, downloading, and running local LLMs with agent-compatible chat completions API. | TypeScript/C++ | Local Inference | App/Web | Link |
| llama.cpp | Minimal C/C++ inference engine for running LLaMA models on consumer hardware with GGUF quantization. | C/C++ | Local Inference | CLI | Link |
| Hivekeep | Self-hosted platform of persistent AI agents that collaborate, build their own tools, and answer over Telegram, Slack, Discord, and Matrix. | TypeScript/Bun/SQLite | Hybrid | Docker | Link |
| Orkas | Local-first desktop workspace where a Commander decomposes goals and coordinates specialist agents; model calls go only to user-selected providers. | TypeScript/Electron | Hybrid | App | Link |
Context architecture, design manuals, and additional directories.
| Repository | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| Prompt Engineering Guide | Exhaustive techniques for structuring LLMs and agents. | Markdown | N/A | Document | Link |
| 12-Factor Agents | Technical specification regarding agentic state control. | Markdown | N/A | Document | Link |
| Awesome LLM Skills | Reusable functions and dependencies for pipelines. | Markdown | N/A | Document | Link |
| Useful AI Prompts | Technical prompts aimed at programmers and deployment. | Markdown | N/A | Document | Link |
| AI Agent Best Practices | Architecture and risk control of scalable systems. | Markdown | N/A | Document | Link |
| Awesome AI Agents | Massive E2B directory with 100+ extra repositories. | Markdown | N/A | Document | Link |
| Awesome Agents (Kyrolabs) | Extensive compilation of autonomy tools. | Markdown | N/A | Document | Link |
| Awesome AI Agents 2026 | List of emerging repositories and trends. | Markdown | N/A | Document | Link |
| Awesome AI Agents (ianwieds) | Categorized list of agent frameworks, coding and browser agents, voice, sandboxes and research. | Markdown | N/A | Document | Link |
Agents integrated into enterprise pipelines, DevSecOps workflows, and corporate infrastructure.
| Project | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| GitLab Duo Agent Platform | Agentic ecosystem integration in native CI/CD cycles. | Integration | API-based | Web UI | Link |
| COSMOS CI Agent | Intelligent dependency and trace analyzer in GitLab repositories. | Python | API-based | Docker | Link |
| Virtual Company Kernel | Corporate structural role simulation with pure AI. | Python | Hybrid | CLI | Link |
| Ray Agentic Core | Base layer for distributed routing, state management, and inference execution across multi-agent node clusters. | C++/Python | Hybrid | Docker | Link |
| Bloop AI | Real-time active cloud container debugger and codebase navigational agent for distributed platforms. | TS/Rust | API-based | Web UI | Link |
Portals to compare frameworks and hosted agents.
| Repository | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| AI Agents Directory | Categorized index of +1,300 utilities. | Web | N/A | Web UI | Link |
| AI Agents List | Feature and pricing evaluation of +600 software solutions. | Web | N/A | Web UI | Link |
| Hugging Face Agents | Native integration and validated list of autonomous models. | Web | N/A | Web UI | Link |
| Glama MCP Directory | Curated directory of MCP servers and AI agent tools with real-time availability checks. | Web | N/A | Web UI | Link |
| Smithery | Agent tool registry and marketplace enabling one-click installation of MCP-compatible agent extensions. | Web | N/A | Web UI | Link |
Optimized semantic processing for professional and regulated fields.
| Project | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| FinRobot | Ingestion and quantitative analysis of corporate documents. | Python | API-based | Library | Link |
| Dexter | Iterative balance sheet research via reflection chains. | Python | API-based | CLI | Link |
| TradingAgents | Distributed environment for transaction simulation. | Python | Hybrid | Docker | Link |
| Project | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| AI Agents Healthcare | Modeling for FHIR compliance and US revenue analysis. | Various | API-based | Library | Link |
| OpenClaw Medical Skills | Pre-trained operations on biomedical variables. | Python | Hybrid | Web UI | Link |
| HealthcareAgent | Unification of logic layers and interfaces in clinical applications. | Python/Java | API-based | Library | Link |
| CareJobs.Now MCP | Remote MCP server for US healthcare job search: open jobs, posted pay, employers and states. | TypeScript | N/A | Library | Link |
| Project | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| LawGlance | RAG engine bounded to regulatory processing. | Python/React | Hybrid | Web UI | Link |
| Octochains | Strict execution conditionals based on regulatory risks. | TS/Python | API-based | Web UI | Link |
| LawAgent | Sub-routines for iterative structuring of legal opinions. | Python | API-based | Web UI | Link |
| Project | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| LogNorm | Hosted MCP server that hands a site's SEO and AI-visibility backlog (audits, fixes, content drafts) to Claude Code, Codex and Cursor. | MCP (hosted) | API-based | CLI | Link |
| NotFair | Claude Code skill set for SEO, GEO, Google Ads, and Meta Ads. Connects to live account data via Google Ads MCP, Meta Ads MCP, Google Search Console MCP, and Google Analytics (GA4) MCP to execute audits, keyword research, creative analysis, and bid management. | TypeScript | API-based | CLI | Link |
| Robot Speed | Hosted SEO and AI-visibility MCP. OAuth remote exposes 39 tools; a rate-limited, read-only no-auth endpoint exposes 12 (docs). EU-hosted (Supabase + Vercel); read-only except 4 write tools. Submitted by the publisher. | Remote MCP (MIT) | API-based | Web UI | Link |
Semantic hypervectorial storage persisting across context limitations.
| Repository | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| Mem0 | Automatic interaction extraction into universal LTM layers. | Python | Hybrid | Web UI | Link |
| MemGPT | RAM block management in LLMs acting as structured OS memory. | Python | Hybrid | Web UI | Link |
| Zep | Asynchronous collector and heuristic router for short-term memory. | Go/Docker | Hybrid | Docker | Link |
| Graphiti | Entity compilation into dynamic graph and relation meshes. | Python | API-based | Library | Link |
| AgentMemory | Indexable segmented vectors focused on code dependencies. | Python | Local Inference | CLI | Link |
| Tree Ring Memory | Local-first memory lifecycle layer with scoped recall, forgetting, and consolidation. | Rust/SQLite | N/A | CLI | Link |
| SoloPup Memory (public beta) | DeepSeek Harness text-memory plugin with source references, project/global scope, conditions and correction history; lexical recall. | Node.js/JavaScript/SQLite | API-based | Library | Link |
Parameterized sets of cases for deterministic measurement.
| Benchmark | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| GAIA | Asynchronous resolution metric via general agents. | Python/HF | N/A | Document | Link |
| SWE-bench | Success ratio on repository merges and bug fixes. | Docker/Py | N/A | CLI | Link |
| AgentBench | Synthetic agent verification across 8 types of sandbox environments. | Python | N/A | CLI | Link |
| WebArena | Automated interactivity score across 4 cloned web microservices. | Docker/TS | N/A | CLI | Link |
| ClawBench | Live-site browser-agent benchmark with 283 V1/V2 tasks and five-layer execution traces. | Python/Docker | N/A | CLI | Link |
| WebsiteBench | Benchmark for agents that reconstruct full-stack websites through browser exploration, evaluated on appearance and behavior. | Python/Harbor/Docker | N/A | CLI | Link |
| YYLO Benchmark | Every candidate receives a dedicated fresh repository; evaluator profiles support deterministic commands and configurable LLM judges, and retained receipts, manifests, terminals, and evaluator provenance form one hash-verified evidence chain. | TypeScript/Node | N/A | CLI | Link |
Experimentation on autonomous group behaviors by unifying multiple LLMs.
| Repository | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| Generative Agents | 2D social emulation based on cyclic priorities over memories. | Python | API-based | Web UI | Link |
| AgentSociety | C++ orchestration for simulating metropolises of up to 30K agents. | C++/Python | Hybrid | Web UI | Link |
| OASIS | Optimized inference for virtual networks of one million entities. | Python/GPU | Hybrid | Web UI | Link |
| Gamma-World | Generative computation and regression of social and mundane variables. | Python | API-based | Web UI | Link |
Frameworks and pipelines designed to enforce policy constraints, prevent prompt injections, and align autonomous agents to specific safety protocols.
| Repository | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| NeMo Guardrails | Configurable execution constraint mechanisms enforcing topological programmable limits via semantic similarity vectors. | Python | Hybrid | Library | Link |
| Llama Guard | Tuned safeguard classifier determining adversarial input topologies to reject unsafe conversational payloads. | Llama | API-based | Library | Link |
| Garak | Systematic vulnerability scanner probing for hallucination vectors and jailbreak surfaces in LLM agents. | Python | API-based | CLI | Link |
| Lakera Guard | Comprehensive API-based security middleware blocking active execution attacks on agentic APIs. | Python | API-based | API | Link |
| Ironheights | Static pre-install scanner for agent skill folders (SKILL.md and bundled files) that flags prompt injection, hidden text, credential reads, remote-exec installs and persistence, with integrity baselines and SARIF output. Rules run offline; optional LLM review via Ollama or an OpenAI-compatible API. | TypeScript/Node.js | Hybrid | CLI | Link |
| HOL Guard | Local-first policy and approval controls for AI coding agents, evaluating shell, secrets and MCP changes before execution. | Python | N/A | CLI | Link |
Small language models (SLMs) and pruned tensors optimized natively for device-edge execution on resource-constrained hardware.
| Repository | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| MLC LLM | Universal compiler deploying LLMs with aggressive static memory allocation to iOS, Android, and WebGPU architectures. | C++/Python | Local Inference | App/Web | Link |
| ExecuTorch | PyTorch-native runtime supporting optimized quantized execution of specialized sub-agents on microcontroller environments. | C++/PyTorch | Local Inference | Embedded | Link |
| TinyLlama | Highly dense 1.1B parameter checkpoint aggressively optimized for near-instant inference routing on portable devices. | Weights | Local Inference | Library | Link |
Autonomous financial agents interacting natively with blockchain nodes, executing deterministic EVM smart contracts, or trading autonomously.
| Repository | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| Agent Payment Node | Local CLI and stdio MCP payments on macOS Apple Silicon; every signature requires foreground owner approval. Only documented EVM/Solana/TRON networks and listed assets are supported. Asset/issuer, chain/provider, irreversible-transfer and variable network/third-party fee risks remain. Payment execution was not tested. | TypeScript / Node 24 | N/A | CLI | Link |
| Eliza | Specialized multi-agent framework tailored for interacting with crypto endpoints, executing DeFi operations, and orchestrating decentralized DAOs. | TypeScript | Hybrid | Node | Link |
| Rig | Rust-native ecosystem building secure Web3 agents capable of deterministic contract execution and complex state transitions. | Rust | Hybrid | Node | Link |
| Goat | The Great Open Agent Toolkit. Middleware connecting off-chain AI reasoning with on-chain EVM transactional networks. | TypeScript | Hybrid | Web UI | Link |
| Spectral | On-chain autonomous agent framework executing DeFi strategies with verifiable transaction logs. | Solidity/Python | Hybrid | Web UI | Link |
| OLAS | Open-source framework for deploying autonomous off-chain agents that interact with smart contracts and DAOs. | Python | Hybrid | Docker | Link |
Frameworks and pipelines optimized for low-latency voice-to-voice communication, streaming speech recognition, and synthesized audio reasoning.
| Repository | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| Pipecat | Open-source framework for building voice and multimodal conversational agents featuring WebRTC bindings and streaming VAD. | Python/TS | API-based | Web/Node | Link |
| LiveKit Agents | High-performance infrastructure explicitly designed to host real-time programmable audio agents over globally distributed WebRTC meshes. | Go/Python | API-based | SDK | Link |
| Vapi | End-to-end voice AI platform abstraction allowing integration of interrupting, context-aware audio agents natively via SIP/WebRTC. | TypeScript | API-based | Web/App | Link |
| OmniChat | Real-time voice agent framework supporting full-duplex audio with interruption handling and emotion detection. | Python | API-based | Docker | Link |
| Retell AI | Voice agent API with turn-taking logic, keyword detection, and custom voice cloning for agent personas. | TypeScript | API-based | API | Link |
Automated systems that ingest chaotic data environments, generate complex SQL/Python transformations, and auto-orchestrate ETL pipelines.
| Repository | Description | Stack | Engine | Deployment | Link |
|---|---|---|---|---|---|
| Dataherald | Natural Language-to-SQL engine constructed over relational databases, supporting RAG-style schema alignment and determinism. | Python | Hybrid | Web UI | Link |
| PandasAI | Conversational data analysis library that allows users to interact with their databases and dataframes using natural language. | Python | API-based | Jupyter | Link |
| LlamaIndex | Though a data framework, its Agentic Data Engines autonomously route, synthesis, and query disparate vector and SQL nodes. | Python/TS | Hybrid | Library | Link |
| Datamatic | Autonomous ETL agent that inspects data schemas, generates transformation code, and validates pipeline outputs. | Python | API-based | Web UI | Link |
| BlazeSQL | Natural language to SQL agent supporting multiple database engines with query optimization and schema learning. | Python | API-based | Web UI | Link |
| AI for Database | Agentic AI that connects to any database (Postgres, MySQL, MongoDB) and enables natural language querying, self-refreshing dashboards, and workflow automation. | SaaS | Hybrid | Web UI | Link |
Contributions are welcome! Please read the Contributing Guide for format requirements and submission guidelines.
Quick start: Fork → Add tool to correct section → Verify link → Submit PR
Entries are reviewed for relevance, link validity, and safe authorized use. Links and projects can change after review; verify a project’s current ownership, license, and security posture before using it. Report compromised or malicious entries privately. See SECURITY.md and CONTRIBUTING.md.