The tools we
actually build with.

Anthropic's flagship LLM family. Strong reasoning, long context, low hallucination rate.

The GPT family. Widest ecosystem, mature tooling, highest cost at production scale.

Google DeepMind's multimodal frontier family. Deep Google Cloud integration.

xAI's frontier model. Fast iteration, tight X-platform integration.

Enterprise-focused frontier models with strong retrieval and multilingual capability. Self-hostable at scale.

Hybrid Mamba-Transformer architecture. Extremely long context with efficient throughput.

Native multimodal models with image, video, and audio understanding built in.

Meta's open-weight family. Self-hostable, commercially usable, community-backed.

European open-weight models. Strong performance-to-size ratios, commercial and Apache options.

Chinese frontier-class open-weight models. Aggressive price-performance, MoE architectures.

Alibaba's multilingual model family. Strong non-English performance, permissive licensing.

Google's open-weight model family. Built from the same research as Gemini, runnable on consumer hardware.

Microsoft's small, capable open models. Tuned for on-device and edge workloads.

NVIDIA's open model family. Strong for building agentic systems on NVIDIA infrastructure.

OpenAI's flagship web interface. Still the default on-ramp for most people using AI.

Anthropic's web client for Claude. Projects, Artifacts, and file-first workflows.

Google's consumer chat surface for Gemini. Deep integration with Workspace and Search.

xAI's chat interface. Real-time access to the X firehose for current-events queries.

AI-powered search with inline citations. A Google replacement for many technical queries.

Quora's multi-model chat client. One interface, many models, useful for comparison.

Native Claude app. First-class MCP server support. The reference client for local tool access.

Native ChatGPT app. Tighter OS integration for voice and screen capture workflows.

Desktop engine for running open-weight models locally. Chat UI + server in one app.

Desktop chat that handles both local and cloud models in one unified interface.

Anthropic's agentic CLI. Autonomously edits code, runs commands, completes multi-step tasks.

OpenAI's open source coding agent. Runs locally with GPT models for code generation and edits.

Google's open source terminal agent for Gemini. Code, reasoning, and tool use from the CLI.

Git-aware pair-programming CLI. Model-agnostic, strong for diff-based edits to large codebases.

Copilot in the terminal. Suggests shell commands and explains existing ones.

Block's open source CLI agent. MCP-native, extensible, strong for developer automation.

The original inline code assistant. Chat, edits, agents, and the widest IDE coverage.

AI-native editor with multi-file edits, agents, and tight inline context. Default for many teams.

Codeium's agentic IDE. Strong Cascade agent for multi-file and cross-repo work.

Open source agentic coding extension. Executes commands, edits files, runs checks inline.

Open source Copilot alternative. Bring your own model, full control over context and rules.

Rust-built editor with first-class AI assistant, inline prediction, and agent panel.

Sourcegraph-powered AI with deep codebase context through their indexing infrastructure.

Native AI across IntelliJ, PyCharm, WebStorm, and the rest of the JetBrains family.

Direct Claude access. Best-in-class reasoning for governed production workloads.

GPT + Whisper + embeddings + fine-tuning. The most mature LLM API on the market.

Unified API for Claude, Llama, Mistral, Titan. Sits cleanly inside existing AWS governance.

OpenAI models on Microsoft's enterprise compliance and identity stack.

Gemini + third-party models with full GCP integration. Strong MLOps tooling.

Unified API across every major provider. Swap models without changing integration code.

Custom LPU hardware for extremely low-latency open-model inference.

Optimized hosting for open-weight models with aggressive throughput and fine-tuning support.

Hosted open-weight models with fine-tuning. Cheap scale for non-frontier workloads.

One-click hosted inference for open models. Strong for image and audio workloads.

Serverless inference for open-weight models. Pay-per-token with no capacity commitment.

Run open-weight LLMs locally. Fastest path from zero to a working private-model environment.

Desktop GUI for local model testing. Useful for evaluation and private workflows.

C++ inference engine underneath most local LLM tooling. Runs on anything that boots.

High-throughput production inference server with PagedAttention. Industry default for self-hosting.

Hugging Face's production serving stack. First-class support for any HF model.

Stateful, graph-based orchestration. Fine-grained control over how agents reason and act.

The original LLM app framework. Chains, retrievers, tools, memory. The batteries-included path.

Data framework for RAG and agents. Strong for document-heavy and knowledge-graph workflows.

TypeScript-first SDK for building AI UX. Streaming, tools, agents, and provider abstraction.

Microsoft's orchestration SDK. First-class across C#, Python, and Java.

Declarative prompting. Compile programs, not prompts. Optimize end-to-end against evals.

Open source framework for search, RAG, and agent pipelines with strong pre-built components.

Open source drag-and-drop builder for LLM flows. Fast prototyping, deployable API endpoints.

Visual LangChain builder. Good fit when non-engineers need to collaborate on agent flows.

General-purpose workflow automation with strong AI node support. Self-hostable.

Anthropic's agent SDK. Production primitives for building and deploying Claude-powered agents.

OpenAI's official agent framework. Handoffs, tool use, guardrails, tracing. Successor to Swarm.

Microsoft's cross-platform SDK for building agents. Successor to Bot Framework, works across Teams, Copilot, web.

SDK for building agents that plug into Copilot Studio and the broader Microsoft 365 Copilot ecosystem.

Google's open source multi-agent framework. First-class support for A2A protocol and Gemini.

Role-based agent teams that collaborate on complex tasks. Fast path to useful multi-agent setups.

Microsoft Research's multi-agent framework. Strong academic and production lineage.

Type-safe agent framework from the Pydantic team. Production-ready, FastAPI-like ergonomics.

TypeScript-native agent framework. Strong for building agents alongside web apps in the JS ecosystem.

Formerly MemGPT. Stateful agents with explicit long-term memory and self-editing context.

Minimalist agent framework from Hugging Face. Code-first agents with small dependency footprint.

Formerly OpenDevin. Fully autonomous software engineering agent running in a sandboxed environment.

Python framework for building agents with structured memory, tool use, and workflows.

Formerly Phidata. Lightweight agent framework with Python-native ergonomics.

Block's open source CLI agent. Extensible through MCP, strong for developer workflows.

AWS-backed open source agent framework. Model-driven loop, first-class Bedrock and AWS integration.

Ontology-anchored AI decision platform. LLMs with governed access to enterprise data, logic, and actions.

The operational system beneath AIP. Integration, ontology, and workflow layer for the enterprise.

AI built on top of the lakehouse. Strong fit when data already lives in Databricks.

Open standard for giving AI governed, contextual access to tools, data, and systems.

The original tool-use API. Still the widest-deployed spec for LLM function invocation.

Agent-to-Agent communication protocol. Emerging standard for agent-ecosystem interop.

Not AI-specific, but the lingua franca for describing APIs that agents need to call.

Discord and web-native image generation. Still the aesthetic benchmark for stylized output.

OpenAI's image model. Tight integration with ChatGPT, strong at prompt comprehension.

The open-weight foundation for most self-hosted image workflows. SDXL, SD3, etc.

Black Forest Labs. Currently the strongest open-weight image model. Built by the Stable Diffusion team.

Google DeepMind's photorealism-focused image family, available through Vertex AI.

Best-in-class for in-image typography and coherent text rendering.

Adobe's commercially safe generative stack. Trained on licensed content, deep Creative Cloud integration.

Image-focused workflow platform with fine-tuning, canvas editing, and production controls.

Gen-3 and Gen-4 video models with a professional editing workflow. Used in actual commercial production.

DeepMind's text-to-video model. Veo 3 delivers high-fidelity output with native audio generation.

OpenAI's text and image-to-video model. Strong for longer, coherent scenes with complex motion.

Fast, playful text-to-video with strong effects library. Popular for short-form and social content.

Chinese text-to-video model with strong realism and long clip support.

Text and image-to-video with fast iteration. Good at camera motion and scene dynamics.

MiniMax's video model. Strong character consistency, competitive on cost per clip.

End-to-end song generation with vocals and instruments. The most capable music model available.

Competitor to Suno with strong audio fidelity and genre flexibility.

Industry-standard voice synthesis and cloning. Production-ready voice agents and dubbing.

Stability AI's audio generation. Open weights available, strong for sound design and loops.

DeepMind's music model. Available through MusicFX and Vertex AI.

Meta's open-weight music generation model. Self-hostable for private workflows.

Meta's open-weight ambient and effects audio model. Environmental sound from prompts.

Text and image to 3D mesh generation. Game-dev and visualization-ready outputs.

Luma's 3D generator. Strong for 3D assets from photos or text prompts.

High-fidelity 3D generation focused on characters and organic models.

AI-generated 360° skyboxes and environments. Used widely in game and VR workflows.

Fast text and image-to-3D generation with PBR materials for real-time rendering.

Open source node-based workflow builder for Stable Diffusion and beyond. Industry standard for serious image pipelines.

The original Stable Diffusion web UI. Massive extension ecosystem, still widely deployed.

Real-time image generation with a canvas-native creative workflow. Strong for iteration.

Unified interface across multiple gen models. Good for comparing outputs quickly.

Cinematic video generation with camera-motion presets. Tuned for motion-design work.

Managed vector database for RAG. Low-latency similarity search at production scale.

Vector database with hybrid search, schema-awareness, and graph capabilities.

Rust-based vector engine. Fast, lightweight, strong for self-hosted deployments.

Dev-friendly open source vector database. Zero-config start, embeddable.

Cloud-native vector database built for billion-scale workloads.

Embedded serverless vector database. Columnar, versioned, strong for multimodal workloads.

Vector search extension for Postgres. Often the right answer when Postgres is already in play.

Vector search in Redis. Useful when Redis is already the cache or session store in your stack.

Native vector search in MongoDB Atlas. Best fit when document data is already in Mongo.

Dense vector + hybrid search in Elasticsearch. Strong when combined with full-text at scale.

The industry standard graph database. Strong tooling, mature ecosystem, Cypher query language.

High-performance analytics graph. Strong for deep-link analysis and fraud detection.

Graph, document, and key-value in one engine. Useful when graph is part of a broader pattern.

Cloud data warehouse with strong data sharing, governance, and now native AI functions.

Unified analytics and AI platform on top of open storage formats (Delta, Iceberg).

SQL-based data transformation with version control and tests. The default for modern ELT.

Open source data ingestion. Large connector catalog, self-hostable.

The streaming event backbone most production architectures eventually need.

Open table format for large analytic datasets. The neutral ground between warehouses.

Enterprise data catalog and governance platform. Default in regulated industries.

Modern data catalog with strong lineage, collaboration, and API-first design.

Open standard for collecting data lineage from pipelines. Vendor-neutral.

Databricks-native governance layer. Fine-grained access control across data and AI assets.

The broadest cloud surface area. Strong for heterogeneous enterprise architectures.

Default cloud for Microsoft-aligned enterprises. Tight identity and M365 integration.

Strongest data and AI primitives. Gemini, BigQuery, Vertex AI all first-class.

Fits Oracle-heavy shops. Strong database integration and competitive egress pricing.

WatsonX AI and regulated-industry positioning. Hybrid and mainframe-adjacent workloads.

On-demand and serverless GPU compute. One of the cheapest paths to run custom model workloads.

Optimized inference for generative models. Sub-second image and video generation at API scale.

Serverless Python functions with GPU support. Clean path from notebook to production.

Model deployment platform. Strong for serving custom and fine-tuned open-weight models.

GPU-specialized cloud infrastructure. Favored for large-scale training and inference.

GPU cloud for training and inference. Competitive pricing, direct NVIDIA partnerships.

Container orchestration. The substrate most production AI workloads end up running on.

Red Hat's enterprise Kubernetes distribution. Default in regulated and hybrid-cloud shops.

Container runtime and image format. Still the baseline for packaging workloads.

Infrastructure as code. The baseline for reproducible, version-controlled cloud environments.

Agentless configuration management. Still the fastest way to automate existing systems.

Declarative, Git-driven continuous deployment for Kubernetes.

Package manager for Kubernetes. Charts are how most production apps get templated.

OpenTelemetry-native open source APM. Traces, metrics, and logs in one UI. Self-hostable alternative to Datadog.

Industry-standard SaaS observability. Broad coverage across infrastructure, APM, logs, and security.

Long-established APM and full-stack observability. Consumption-based pricing across telemetry types.

The de facto open source stack for metrics, dashboards, and alerting. Runs anywhere.

Event-oriented observability for complex distributed systems. Strong for debugging production unknowns.

APM inside the Elastic stack. Fits naturally when logs and search already live in Elasticsearch.

AI-powered enterprise observability. Strong auto-instrumentation and root-cause analysis.

Error tracking and performance monitoring focused on application code. The default for front-end and app errors.

The vendor-neutral telemetry standard. Instrument once, export anywhere. Backbone of modern observability.

LangChain's observability and evaluation platform. Tight integration with their stack.

Open source LLM engineering platform. Tracing, evals, prompt management.

Evaluation-first platform for AI products. Strong for iteration speed on production prompts.

Open source observability for LLMs. Visual traces, evaluations, datasets.

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

Formal modeling of business objects, relationships, and logic. The anchor for every downstream AI system.

LLM responses grounded in retrieved context from your own data. The baseline pattern for enterprise AI.

Multi-step reasoning where AI plans, calls tools, evaluates results, and iterates. Beyond single-shot prompts.

AI proposes, humans approve, systems execute. Accountability without slowing the flywheel.

Access scoped by role, data classification, and intent. Not just who. Also why.

Systematic measurement of model behavior over time. The difference between production AI and a hopeful pilot.

Automatic tagging of data by type, sensitivity, and meaning. Makes governance scale past manual review.

Git as the source of truth for infrastructure and deployment. Auditable, reproducible, rollback-able.

Relational representation of entities and their links. Gives AI structural context beyond bag-of-words retrieval.

Wrapping existing business logic, APIs, and data as callable tools for AI agents through MCP servers.

A diagnostic framing: fine-tune for style and format, retrieve for facts. Conflating them burns budget.

Deterministic checks around non-deterministic models. Input filtering, output validation, scope enforcement.

How Palantir frames the decision platform: ontology, tools, actions, scenarios, guardrails.

The Anthropic-authored open standard for AI-to-system integration. Read it before you build.

Practical guidance from the team building Claude. Strong on agents and evaluation patterns.

Practical, runnable examples from OpenAI. Still the best source for day-one integration patterns.