New in week 43

Database for AI with a deep-learning storage format for embeddings, media and datasets, plus vector search.

Open source gateway for HTTP, gRPC, LLM, MCP and agent-to-agent traffic in one data plane.

Open-source, datacenter-scale orchestration layer that coordinates multi-node inference across engines such as SGLang, TensorRT-LLM and vLLM.

Open-source API gateway with AI plugins for multi-provider LLM proxying, token limits and observability.

Web scraping and browser automation library for Node.js, covering link discovery, data extraction and storage.

Small SQLite extension that adds vector search to SQLite tables, written in pure C with no dependencies.

AWS platform for deploying and operating production AI agents with any framework and any model.

Inference and training platform for hosted and dedicated models, deployable in your cloud or Baseten's.

Open-source, Markdown-based knowledge graph memory that AI clients use over MCP.

CPU and GPU machines and sandboxes for agents, billed by the millisecond, with VMs that pause with memory intact.

Unified inference platform for deploying and scaling AI models on any cloud, with an open-source Python serving framework.

Open-source AI gateway with a Python SDK that calls 100+ LLM providers through the OpenAI format.

Open-source SDK for browser agents that combines natural-language act, extract and observe with Playwright-style control.

Headless browsers that run on Browserless cloud or in your own Docker deployment, reached over WebSocket and REST.

Open-source data infrastructure for AI that stores embeddings and runs vector, full-text and metadata search.

Durable runtime for building and hosting AI agents on Cloudflare's global network.

Sandboxes on Cloudflare Workers that run untrusted or agent-written code in isolated containers.

Continuously fresh context for AI agents, with only changed data reprocessed.

Platform to make AI agents and workflows durable, deterministic and explainable.

Python framework for LLM-powered agents, built on the Dapr distributed systems runtime.

Open-source tool that converts PDFs, images and Office files to markdown, JSON, chunks and HTML.

Secure, elastic infrastructure that runs AI-generated code in sandboxes created in under 90 milliseconds.

Durable workflows and queues that run on Postgres, for backends and AI agents.

Docker's open source MCP Gateway that runs MCP servers in isolated containers with credentials and access control.

Open-source document conversion library that turns PDFs, Office files and audio into structured data.

Open-source infrastructure that runs AI-generated code in isolated cloud sandboxes.

Elasticsearch as a vector database, combining vector search with hybrid lexical search, filters and reranking.

Library for efficient similarity search and clustering of dense vectors, with optional GPU support.

API to search, scrape and interact with the web for AI agents, open source and available as a hosted service.

Multimodal LLM that takes speech input directly, without a separate speech recognition stage, for real-time voice agents.

Open-source Python framework for building temporal knowledge graphs that AI agents can query.

Plain C/C++ LLM inference engine that runs quantized models on many hardware backends, with an OpenAI-compatible server.

Durable runtime that runs AI agents and their tool calls in WebAssembly VMs that resume after crashes.

Orchestration platform for AI agents, background tasks and durable workflows.

Simple and fast retrieval-augmented generation library with several chunking strategies and storage backends.

Open source registry and proxy that federates MCP, A2A and REST/gRPC APIs under central governance.

Open-source RAG engine with document ingestion, hybrid search and visual agent workflows.

Open-source durable execution platform for orchestrating async code, background jobs and agents.

Converts any URL into LLM-friendly input and searches the web, through the r.jina.ai and s.jina.ai endpoints.

AI-native proxy and data plane for agent apps: agent orchestration, LLM routing, guardrails and traces.

Open-source browser infrastructure that runs each Chromium browser in its own VM for agents, with fast starts.

Open-source orchestration platform for data, AI and infrastructure workflows, defined in YAML.

Cloud-native API, LLM and MCP gateway with multi-LLM routing, AI analytics and an extensible plugin system.

Kubernetes CRD and controller for isolated, stateful sandbox pods that behave like single-container VMs.

Open-source, embedded vector search and multimodal lakehouse built on the Lance columnar format.

Library that lets LangGraph agents extract, consolidate and search their own long-term memory.

Platform for stateful agents that learn from experience and keep their memory across sessions.

Headless browser built from scratch in Zig for AI agents and automation, not based on Chromium.

Open-source document intelligence API that converts PDFs, Office files and images into RAG-ready chunks.

Open-source AI gateway with one OpenAI-compatible API for 23+ providers, with fallbacks and semantic caching.