Find the AI app for the job.

Cascade agent + IDE. Codeium's developer surface.

Agent-first IDE that pairs autocomplete with Cascade, a planning-then-acting agent. Strong terminal integration; good when the diff spans many files.

Hosted eval + tracing platform for LLM apps.

Production-grade eval orchestration with a dashboard, dataset versioning, and OpenTelemetry tracing. Useful once eval volume outgrows a CI YAML file.

Long-term memory layer for AI agents. Self-hostable.

Persistent memory store + retrieval pipeline for agent applications. Handles per-user/per-session/per-agent scope cleanly; pairs with OpenAI, Anthropic, and local models.

Stateful agents with structured memory. Successor to MemGPT.

Open-source framework for building stateful agents — memory blocks, context-window management, tool-use primitives baked in. Useful as a reference architecture for long-running agents.

Autonomous software engineer agent. Plans, codes, ships from a prompt.

Cognition's hosted SWE agent. Operates inside its own sandbox with a browser, terminal, and editor — plans the work, writes the code, and submits a PR. Aimed at long-horizon tickets rather than tab-completions.

General-purpose autonomous agent for research, ops, and content.

Cloud agent that browses, drafts, schedules, and follows multi-step instructions over hours. Closer to a virtual operator than a coding tool — strong for non-code knowledge work. Acquired by Meta; product continues to ship.

Open-source autonomous SWE agent. Successor to OpenDevin.

Self-hostable agent platform with a browser, terminal, and editor in a sandbox — comparable to Devin in shape, open in licence. Active research community pushing the autonomous-SWE bar.

Prompt-to-app generator running in the browser. WebContainers under the hood.

StackBlitz's in-browser full-stack app builder. Generates a working project — frontend, backend, deps — and lets you iterate in a sandboxed Node runtime without leaving the tab.

Prompt-to-UI generator from Vercel. Outputs React + Tailwind + shadcn.

Vercel's generative UI surface. Type a description, get a working React component using Tailwind and shadcn primitives — copy the code or fork into a v0 project for iteration.

Graph-based agent orchestration. Stateful loops with checkpoints.

LangChain's agent layer. Model agents as nodes in a state graph with persistent checkpoints, human-in-the-loop steps, and durable execution. Strong when you need a long-running, debuggable agent rather than a one-shot chain.

Multi-agent framework with explicit roles and tasks.

Python framework for orchestrating crews of specialised agents — researcher, writer, reviewer — coordinated through shared context. Opinionated about roles, sequencing, and delegation; good fit for content-and-research pipelines.

Open-source workflow automation with first-class AI nodes.

Visual workflow builder bridging APIs, databases, and AI providers. Self-hostable; commercial cloud available. The default Zapier-style surface for the agentic-workflow crowd.

Durable workflow engine for AI background jobs.

Event-driven, durable execution engine — pause/resume, retries, fan-out, scheduling — designed for long-running AI jobs that can't sit in a request/response cycle. TypeScript-first; framework-agnostic.

Open-source vector database with built-in vectorisers.

Cloud-native vector DB that can compute embeddings inline — pass raw text in, store vectors out. Strong hybrid (BM25 + vector) search; self-hostable or managed.

Open-source, Rust-based vector DB. Fast, predictable, self-hostable.

Vector database written in Rust with a strong focus on filtering, payloads, and predictable latency at scale. Self-host on a single binary or use the managed cloud.

Embedded vector DB. Pip-install, prototype, scale later.

The low-friction starting point — Chroma runs embedded inside your Python process or as a hosted service. Great for prototypes and small-to-medium RAG apps; upgrade to a managed option when you outgrow it.

Object-storage-backed vector DB. Serverless economics at scale.

Bills like S3 — cold rest, warm reads, no per-namespace minimums. Designed for very-large, mostly-cold vector workloads where you can't justify keeping every index in RAM. Operated by Notion in production.

Voice agent infrastructure. Build a phone-agent in a weekend.

Production voice-agent platform — telephony, STT, LLM, TTS, and interrupt handling stitched together so you call an endpoint and get a working phone agent. Pluggable models at every layer.

Low-latency streaming TTS. Sub-100ms first audio.

Streaming-first speech synthesis built around the Sonic family of state-space models. Aims at real-time agent voices where latency between turns is the product. Strong choice for sub-200ms voice loops.

Production speech-to-text. The STT default for many companies.

End-to-end speech recognition platform — real-time streaming, batch transcription, speaker diarization, and language detection. Strong on accented speech, telephony audio, and long-form recordings.

Visual prompting + vision agents from Andrew Ng's lab.

Build vision applications with a labelling-light workflow — point at examples, get a deployable detector. Recently extended into vision agents that reason over images and PDFs without bespoke training.

FiftyOne — open-source vision data platform.

Open-source toolkit for exploring, debugging, and curating vision datasets. Strong story for finding model failure modes, balancing classes, and tracking experiment drift across visual data at scale.

Image gen with the best text-in-image quality.

Hosted image generator with industry-leading typography rendering — actually readable text inside images. Strong for logos, posters, social cards, and any composition where letterforms matter.

Design-loop image surface. Generate, edit, upscale, iterate.

Multi-model creative canvas — image generation, real-time edits, upscaling, and style transfer chained together. Strong for fast iteration on a single image rather than rolling fresh prompts.

AI upscaling + relighting that goes beyond pixel-doubling.

Upscales images while reimagining detail — adds plausible texture, lighting, and structure rather than just sharpening. Broke through with the 'AI photography' workflow for product shots and concept art.

OpenAI's video generator as a standalone app.

OpenAI's video model surfaced as a consumer product — generate, remix, and share short videos from prompts or images. Available standalone and bundled into ChatGPT for subscribers.

Mainstream-quality video generation. Indie creator default.

Luma Labs' video model surfaced as an iOS app and web tool. Strong text-to-video, image-to-video, and keyframe controls — the indie creator's go-to before reaching for Runway's heavier toolkit.

Avatar video at scale. Talking-head clips from a script.

AI avatar platform for B2B content — generate a presenter from a photo, give them a script, get a finished video with lip-sync, voiceover, and translation. Used heavily in marketing and corporate training.

AI music generation. The Suno alternative.

Hosted full-song generation focused on production-quality output and finer style control. The other half of the AI music duopoly — many producers run prompts through both Suno and Udio.

Podcast + audio editing where the transcript is the timeline.

Audio and video editing built around an editable transcript — cut words, get cut audio. Add AI cleanup, overdub voice clones, and screen-recording for podcasts and tutorials in one tool.

Search API built for AI agents. First-class in most agent frameworks.

Search-as-a-tool for LLM agents — returns scrape-friendly results tuned for retrieval rather than ranking. Native integrations across LangChain, LangGraph, CrewAI, and the major agent surfaces.

Neural search API. Find pages by meaning, not keywords.

Semantic search engine that indexes the open web with embeddings — pass a description, get matching pages. Strong for research-style queries and find-similar workflows; formerly known as Metaphor.

Independent search index. Privacy-respecting, non-Google.

API access to Brave's independently-crawled search index. Useful when you need a non-Google source, want to dodge upstream rate limits, or have privacy posture requirements that rule out scraping competitors.

Drop-in LLM proxy with logging, caching, and cost tracking.

One-line integration — change your OpenAI/Anthropic base URL and get a dashboard with every prompt, response, latency, and dollar tracked. Adds caching and rate-limit handling without code changes.

LangChain's hosted observability + eval platform.

Tracing, dataset management, eval orchestration, and prompt playground from the LangChain team. Pairs naturally if LangChain or LangGraph already runs in your stack, but works standalone via SDKs.

Open-source LLM tracing + evaluation. Strong on retrieval debugging.

Phoenix is Arize's open-source observability — run locally in a notebook or as a service. Especially strong for inspecting RAG pipelines, finding bad chunks, and tracking retrieval quality over time.

Serverless GPUs. Run training, inference, batch jobs from Python.

Define cloud workloads in Python, deploy with one command — GPU access on demand, fast cold starts, fair-share pricing. The default 'I need to fine-tune a model from a Jupyter cell' platform.

Replace frontier-model spend with a fine-tuned small model.

Captures your production OpenAI / Anthropic calls, builds a dataset, fine-tunes a small open-weights model on your traffic, then serves the swap behind your existing SDK. The pitch: 10x cost reduction at parity.

Fast inference + fine-tuning. Production deployments at scale.

Optimized inference platform for open-weights models with strong latency numbers and serverless + dedicated deployment options. Fine-tuning supported; vision and audio models alongside text.

All-in-one in-browser AI assistant. Multi-model.

Same shape as Sider — multi-model AI sidebar that follows you across the web. Strong on quick actions (summarise, rewrite, translate) and saved prompt templates.

Highlight, summarise, and chat with anything you read on the web.

Web-highlighter + AI extension. Save passages from any page, then ask AI for summaries, related ideas, or chat across your knowledge base. BYO key keeps your reading history private.

Visual-canvas automation. The Integromat folks went big on AI nodes.

Drag-and-drop scenario builder with first-class AI provider nodes. More expressive than Zapier for complex routing and conditional logic; favoured when your automation has branches and loops.

Agentic automation. Long-running agents instead of one-shot triggers.

Builds AI agents that wait for emails, calendar invites, or webhooks — then take multi-step actions on your behalf. The 'agentic Zapier' pitch; strong for ops workflows that need judgement, not just routing.