Categories
25 categories. Benchmarks, pricing, and field evidence organised by use case.
SDKs for building custom autonomous agents (LangChain, CrewAI, Mastra, AutoGen).
Autonomous agents that ship code end-to-end (Claude Code, Devin, Aider, Codex).
Long-running, durable AI workflows and job queues (Inngest, Trigger.dev, Temporal).
Prompt-injection defence, safety testing, jailbreak detection (Lakera, Robust Intelligence).
Agentic browser control and web-task execution (Browserbase, Stagehand, Playwright MCP).
Toolkits for retrieval-augmented generation pipelines (LlamaIndex, Haystack, Vectara).
LLM-call tracing, cost tracking, and evaluation logs (Langfuse, Helicone, Honeycomb).
Multi-model routing, failover, and observability layers (OpenRouter, Vercel AI Gateway, Portkey).
In-IDE pair-programmers and inline code suggestion tools (Copilot, Cursor, Continue, Cody).
Local LLM runtimes and on-device inference (Ollama, LM Studio, MLC, Candle).
Image and video generation for product UI and brand work (Midjourney, Ideogram, Leonardo).
Foundation-model APIs (Anthropic, OpenAI, Google, Mistral, etc.).
Vector storage and approximate-nearest-neighbour search (Pinecone, Weaviate, Qdrant, Chroma).
Voice-native AI agents and real-time speech infrastructure (Vapi, Retell, Deepgram Aura).
Managed fine-tuning of open-source and proprietary models (Modal, Together Fine-tuning, Fireworks Fine-tuning).
Semantic code search and large-codebase navigation (Sourcegraph, Greptile, GitHub Code Search).
Third-party hosts for open-source LLMs (Together, Fireworks, Groq, Replicate).
Document parsing, layout understanding, and structured extraction (Reducto, LlamaParse, Unstructured).
Prompt versioning, experimentation, and rollouts (PromptLayer, LangSmith, Pezzo).
Standalone IDEs built around AI workflows (Cursor, Windsurf, Zed).
Train, serve, and deploy custom models (Modal, Replicate, Hugging Face).
AI-assisted code review and PR-comment automation (CodeRabbit, Greptile, Cubic, Sourcery).
Frameworks for benchmarking and regression-testing model outputs (Promptfoo, Braintrust, Patronus).
Embedding model endpoints for semantic search and retrieval (Voyage, Cohere Embed, OpenAI Embed).
Synthetic data generation for training and testing (Tonic, Mostly AI, Gretel).