DecisionBox
DecisionBox is an autonomous warehouse-analysis agent that writes and validates its own SQL, then turns findings into a ranked backlog of insights instead of waiting for a human to ask the right question.
DecisionBox is an autonomous warehouse-analysis agent that writes and validates its own SQL, then turns findings into a ranked backlog of insights instead of waiting for a human to ask the right question.

Agent Browser Shield is a Chromium extension that cleans pages before a browser agent reads them, cutting token waste while masking PII, stripping noisy chrome, and blocking common prompt-injection vectors.

Asteroid is a browser-agent platform for operational teams that need AI to log into portals, move through multi-step workflows, and keep humans in the loop instead of juggling brittle RPA scripts.

Hermes WebUI is a self-hosted browser interface for Hermes Agent that brings chat, files, sessions, memory, and agent controls into a web app without giving up the underlying CLI-style power.

Agentspan is an open-source runtime layer for durable AI agent execution, adding persistence, approvals, memory, and observability without forcing teams to redesign their agents.

Hindsight is an open-source agent memory system that gives AI agents durable, structured memory so they can recall facts, learn from outcomes, and improve across sessions.

OpenSwarm is an open-source multi-agent workspace that turns one prompt into coordinated outputs across research, docs, slides, data, images, and other deliverables.

Envelope is a platform and open standard for defining AI agent teams, versioning how they work, and deploying them across orchestration environments without forcing teams into one proprietary workflow.

OpenAgent is a self-hosted personal AI assistant platform that combines MCP, RAG, browser and computer use, and multi-provider model support in one package.

Open Comet is an open-source browser sidepanel agent for researching pages, taking actions, and running multi-step web workflows with local or BYOK models.

Jarvis Registry is an enterprise MCP and agent gateway for connecting copilots to internal tools through one governed layer with identity, ACLs, discovery, and observability.

AINL is an AI-native language and runtime model for graph-based agent workflows, constraints, and tool-driven execution with less dependence on long prompt loops.

Band is an infrastructure layer for multi-agent systems, giving agents persistent identity, structured memory, coordination, and auditability across frameworks.
mem9 gives coding agents and workflow apps one shared persistent memory layer so context survives across sessions, machines, and different agent runtimes.

iii is a code-first runtime for building APIs, workflows, observability, and AI agents on one shared execution surface instead of stitching every backend component together by hand.
Memory Store syncs conversations, snippets, and decisions from the tools your team already uses, then exposes that context as shared memory for agents and humans through MCP.

Owlish turns websites, FAQs, docs, and PDFs into customer-support agents that can answer, act, cite sources, and hand work off to humans.

TradingAgents is an open-source multi-agent trading framework that coordinates analyst, researcher, trader, and risk roles to study markets with LLM-driven workflows.

OpenHuman is a personal AI agent platform focused on private, powerful assistance for users who want more control than consumer chatbot apps usually provide.

Flowly is a native desktop AI assistant for macOS, Windows, and Linux that can act across apps and browser tabs with chat, voice, and mobile companion workflows.

Poke is a proactive personal AI assistant that lives in messaging apps such as iMessage and WhatsApp and can handle tasks through text.

Perplexity Computer is a 2026 multi-model agent system for long-running research, coding, finance, and workflow execution across cloud and Mac-based environments.

Gemini Spark is Google’s 2026 personal AI agent for background tasks, connected apps, and user-directed 24/7 productivity workflows.

Kimi WebBridge is a local browser bridge for AI agents. It pairs a browser extension with a local service so an agent can click, navigate, fill forms, screenshot, and keep working across sites that do not expose clean APIs.