New in week 43

Long-term memory layer that lets AI assistants and agents keep user context across sessions.

Open-source long-term memory layer for AI agents with episodic and profile memory.

Memory built from what agents do, not only what they say, so later runs can avoid repeated failures.

Memory operating system for LLMs and AI agents that stores, retrieves and manages long-term memory.

Single-file memory layer for AI agents with persistent, versioned and portable memory and no database.

MCP gateway that gives company agents approved tools, role-based access and a record of every call.

Microsoft Research pipeline that builds knowledge graphs from text to improve LLM question answering.

Open-source, high-performance vector database for large-scale search, with a managed option on Zilliz Cloud.

Memory system for agents that captures, structures and retrieves memories, with a Docker backend and Python client.

Unified gateway for deploying and managing multiple LLM providers behind one secure endpoint.

Database for unstructured and multimodal data that powers AI apps and agents, with MCP support.

Single-file executable that bundles llama.cpp with Cosmopolitan Libc, so an LLM runs locally with no installation.

Open-source AI runtime that runs text, voice, image and agent workloads on your own hardware behind OpenAI-compatible APIs.

Browser infrastructure for AI agents: stealth cloud browsers, an open-source Playwright-based framework and deployable functions.

Open-source library that optimizes LLM and visual generation inference with specialized kernels and a Python framework.

Runs open models locally or in the cloud, with a REST API and Python and JavaScript libraries.

Document parsing engine for LLMs and agents, turning PDFs, Office files and web pages into Markdown or JSON.

Open-source sandbox platform that runs AI agent workloads in isolated environments on your own infrastructure.

Open-source vector database inside the OpenSearch search suite, for semantic search and RAG.

Python framework for live data processing and RAG pipelines that run on a Rust engine.

Open-source vector similarity search extension for Postgres that stores vectors alongside relational data.

Open-source memory for AI agents that reasons over conversations to model each user and agent.

AI gateway giving one API to 250+ models, with observability, guardrails and prompt management.

Enterprise MCP gateway and governance layer for discovering, managing and controlling MCP servers.

Durable orchestration for data, ML and agent workflows, written as Python functions.

Open-source vector similarity search engine and database written in Rust, with a managed cloud.

Managed Ray platform for distributed training, batch inference and data processing on GPU clusters.

Managed memory layer from Redis that lets AI agents recall facts and preferences across sessions.

Build vector databases on Redis, with vector search combined with exact-match and full-text search.

Hosted API for running, fine-tuning and deploying AI models, including image, video, speech, music and LLMs.

Open durable execution: write async await code that runs on the database or queue you choose.

Durable runtime for resilient, stateful applications, shipped as one self-contained binary.

Open-source, self-hostable orchestrator for agent workloads built on durable Actors.

AI parsing tool that turns complex documents, tables and charts into clean markdown at scale.

Cloud GPU platform for AI with pods, pay-per-second serverless endpoints and clusters across 31 regions.

Python library and API that uses LLMs to extract structured data from websites and local documents.

Free, self-hostable metasearch engine that aggregates results from many search services without tracking users.

Open-source inference framework for LLMs and multimodal models, built for agentic workloads and large-scale serving.

Rust crawling engine for scraping and crawling at scale, runnable locally or through the Spider Cloud API.

Open-source platform that runs MCP servers in isolated containers, with access policy and observability.

Open-source browser API for AI agents, with sessions, proxies and cloud browsers that you can also run yourself.

Memory and context engine for AI that extracts facts, keeps user profiles and feeds context back to agents.

Open-source, local microVM sandboxes for untrusted workloads, with an embeddable SDK and an MCP server.

Open-source durable execution platform for building reliable applications and AI agents.

Open source control plane for AI and agent traffic, powered by Envoy, with one OpenAI-compatible API.

PostgreSQL extension that builds on pgvector with a DiskANN-style index and label-filtered vector search.

Code sandboxes built from Docker-image snapshots for running agent and model-generated code, with Python and TypeScript SDKs.

Open-source AI memory platform that turns documents, code and conversations into a knowledge graph for agents.

Open-source platform for building durable AI agents and workflows in TypeScript.

Open-source web crawler that turns pages into LLM-ready Markdown, with an optional hosted cloud API.