Harness
Agent harnesses that wrap model calls, tools, and the agent loop
Hermes Agent
Jeffrey Quesnelle
Nous Research's open-source self-hosted agent with persistent memory and self-created skills.
Mastra
Sam Bhagwat
TypeScript agent framework — agents, tools, and durable workflows, with tracing and evals built in.
OpenClaw
Peter Steinberger
Viral open-source personal AI agent that lives in your chat apps.
Pi
Mario Zechner
Minimal open-source coding agent harness, extensible via TypeScript.
Databases
Databases and storage systems agents persist state and knowledge in
Browser Use
browser and computer use
Browserbase
Paul Klein IV
Cloud browsers on demand via API.
Hyperbrowser
Shri Sukhani
Cloud browsers built around the anti-bot layer — stealth fingerprinting, proxy rotation, and CAPTCHA solving — plus scrape and extract endpoints.
Kernel
Catherine Jue
Open-source cloud browser infrastructure for agents, built around fast cold starts and persistent sessions.
Code Sandboxes
Isolated environments where agents run untrusted or generated code.
Credential Vaults
Stores that give agents scoped credentials without exposing secrets to the model.
Human-like Communication
Memory
Mem0
Taranjeet Singh
Memory layer that extracts durable facts from conversations and recalls them across sessions.
Memorable
Nikhil Krishnaswamy
Procedural memory — stores how a task was done as a graph, so agents recall proven procedures instead of re-deriving them.
Supermemory
Dhravya Shah
Memory API bundling retrieval, connectors, and user profiles into one context layer.
Zep
Daniel Chalef
Temporal knowledge graph memory for AI agents.
OAuth and Tools
Managed OAuth and hosted tool layers that connect agents to third-party APIs.
Arcade
Alex Salazar
MCP-native tool runtime with per-user OAuth, where tokens are injected at call time and never reach the model.
Composio
Soham Ganatra
Managed OAuth plus a hosted tool layer connecting agents to third-party apps.
Executor
Rhys Sullivan
Open-source MCP gateway — connect every integration and account once, then any MCP client reaches all of it through one endpoint.
Nango
Robin Guldener
Open-source managed OAuth and unified APIs for product integrations.
Observability and Evals
Tracing, logging, and evaluation for agent and LLM systems.
Arize
Jason Lopatecki
Enterprise ML and LLM observability, with open-source tracing in Phoenix.
Braintrust
Ankur Goyal
Evals, logging, and prompt iteration for LLM products.
Langfuse
Marc Klingen
Open-source tracing, evals, and prompt management for LLM apps.
Raindrop
Zubin Koticha
Sentry for AI agents — custom classifiers that catch silent failures in production, then suggest fixes.
The Context Company
Rohil Agarwal & Arman Kumaraswamy
Agent observability from the user's side — clusters conversations to surface frustration, confusion, and silent task failures.
Payments and Spend Authority
Wallets, cards, and spend controls that let agents pay for things with scoped authority.
Retrieval and Parsing
Turning documents and unstructured files into data an agent can actually use.
Extend
Kushal Byatnal
End-to-end document pipelines — parse, extract, classify, split, and edit — built for production workflows rather than one-off parsing.
LlamaIndex
Jerry Liu
Open-source framework plus LlamaParse — document parsing and indexing that feeds agents over private enterprise data.
Reducto
Adit Abraham
Vision models that turn complex PDFs — tables, figures, handwriting — into LLM-ready structured data.