New in week 43

Platform that turns PDFs, tables, forms and scans into structured, agent-ready data.

Durable serverless functions in TypeScript and Python, with retries, delays and event waits.

Open-source, cloud-native distributed vector search engine for approximate nearest neighbor search.

Agent memory system that helps agents learn over time through retain, recall and reflect operations.

Browser automation CLI for AI agents, built as a native Rust binary.

Isolated Linux microVMs for running agent-generated or untrusted code, driven by an SDK and CLI.

Makes ordinary TypeScript async functions durable, with suspend, resume and retries.

Open-source platform for search, ranking and ML inference over vectors, tensors, text and structured data.

Fast, easy-to-use open-source library for LLM inference and serving, with an OpenAI-compatible server.

Open-source, cloud-native vector database for semantic search, RAG and hybrid queries.

Open-source, code-first platform for scripts, workflows, internal apps and AI agents.

Open-source platform for serving LLM, embedding and multimodal models on your own premises or in the cloud.

Platform that uses LLM prompts to turn unstructured documents into structured JSON, deployable as an API or ETL pipeline.

Managed Google Cloud feature that generates, consolidates and stores long-term memories for agents.

Vector Search engine built on Google Research's ScaNN algorithm for semantic search and recommendations.

Conversational AI engine that lets any LLM listen and reply in real time, with ASR and TTS selection, interruption handling and noise suppression.

Single API to 1,000+ AI models across chat, image, video, audio and voice, billed under one account.

Turns plain-English workflow descriptions into browser agents that run on schedules or triggers.

Cloud platform to build, deploy and distribute MCP servers and MCP apps for ChatGPT and Claude.

Managed, isolated browser environment for agents working in web applications, with live view and session replay.

Managed AWS sandbox where AI agents write and run code, with pre-built language runtimes and file handling.

Fully managed AWS service that gives agents short-term memory within a session and long-term memory across sessions.

Object storage with native vector storage and querying for AI agents, RAG and semantic search.

Cloud browsers for AI agents with managed authentication, natural-language tasks and stealth browsing.

Marketplace of ready-to-run web scraping and automation tools called Actors, runnable through an API.

Apache Cassandra-based NoSQL database with vector search for generative AI workloads.

Azure search service with full-text, vector, hybrid and multimodal retrieval for RAG and agents.

AI gateway capabilities in Azure API Management to secure, scale, monitor and govern AI backends.

Hyper-V isolated code interpreter sessions in Azure Container Apps for running LLM-generated or untrusted code.

Voice AI agents for regulated, high-stakes phone calls, run end to end on Bland's own speech, language and voice models.

Perpetual sandboxes that idle to standby and resume in about 25 milliseconds, with shared agent storage.

Search API for developers and AI agents, built on Brave's own index without relying on Google or Bing.

Hosted browsers for Puppeteer, Playwright and Selenium scripts, with built-in unblocking and CAPTCHA solving.

Pre-built scrapers that return structured data from sites such as LinkedIn, Amazon and Instagram.

Cloud browsers and hosted browser agents that connect over CDP and get past bot checks.

Cloud browsers and web-data APIs for agents, including browser sessions, search, fetch and agent identity.

Context and memory layer that syncs project decisions and rules selectively across coding agents and teammates.

Speech-to-text, text-to-speech and managed voice agents in one API for live conversation.

Fast AI inference cloud on wafer-scale hardware, with an OpenAI-compatible API and a free trial.

Real-time AI infrastructure for deploying voice agents, video models, LLMs and other AI workloads, with autoscaling.

Gateway for AI apps that adds analytics, logging, caching, rate limiting and retries to model calls.

Headless Chrome on Cloudflare's global network for automation, scraping, screenshots and testing.

Globally distributed vector database for building AI applications on Cloudflare Workers.

Serverless GPU inference on Cloudflare's global network, with 50+ open-source models and pay-for-what-you-use pricing.

Durable multi-step applications on Cloudflare Workers that can wait for events or approvals.

Context layer for building AI agents that reason over technical documents and institutional knowledge.

AI cloud platform for GPU compute, training and serverless inference, with Forge tools for the model lifecycle.

AI cloud with GPU instances, managed inference, serverless fine-tuning and Kubernetes or Slurm GPU clusters.

Global WebRTC infrastructure and SDKs for realtime voice and video, from the team behind Pipecat.

One real-time API for voice agents combining speech-to-text, LLM orchestration and text-to-speech, priced at $4.50/hr.