Tooling & Orchestration — SDKs, agents, frameworks
Fable 5 · data curated through July 2026
Layer 4 of five in the Five Layer AI Cake. 24 companies, 21 events.
What this layer is
The connective tissue — how developers wire models into real software.
Layer 4 of five. Below it sits Foundation Models, which it pays. Above it sits Applications, which pays it.
24 of the atlas's 103 companies capture value here. 15 of them also capture value on another layer, which makes them vertical integrators; 9 compete on this layer alone.
Vertical integrators on this layer (15)
These companies span two or more layers, so they hold margin at every floor they occupy.
| Company | Founded | Also on | Influence | Why it is here |
|---|---|---|---|---|
| OpenAI | 2015 | Foundation Models · Applications | 10/10 | ChatGPT, GPT-4, Sora, o-series. Defined the category. |
| Anthropic | 2021 | Foundation Models · Applications | 9/10 | Claude, MCP, constitutional AI. Safety-first frontier lab. |
| Hugging Face | 2016 | Foundation Models | 8/10 | The GitHub of ML. 1M+ models hosted. |
| Meta AI | 2013 | Foundation Models | 8/10 | Llama, PyTorch — the open-weight anchor. |
| Claude Code | 2025 | Applications | 7/10 | Agentic CLI. Killer app for coding agents. |
| Databricks | 2013 | Cloud & Compute · Foundation Models | 7/10 | Acquired MosaicML ($1.3B). Enterprise AI plumbing. |
| Vercel | 2015 | Cloud & Compute | 7/10 | AI SDK, v0, AI Gateway. Frontend-first infra. |
| Qualcomm | 1985 | Silicon | 6/10 | Bought Modular to fight CUDA with software. Silicon without a dev stack does not sell. |
| NVIDIA GR00T · Isaac | 2024 | Foundation Models | 5/10 | Humanoid foundation model platform + Isaac simulation. Robotics dev platform. |
| Replit | 2016 | Applications | 5/10 | Agent IDE in the browser. Solo-founder-friendly. |
| Snowflake | 2012 | Cloud & Compute | 5/10 | Cortex LLM functions. Enterprise data gravity. |
| Together AI | 2022 | Cloud & Compute | 5/10 | Open-model inference cloud. |
| Fireworks AI | 2022 | Cloud & Compute | 4/10 | Low-latency inference for open models. |
| Replicate | 2019 | Cloud & Compute | 4/10 | Run any model via HTTP. |
| Paperclip | 2025 | Applications | 3/10 | Open-source AI labor management — agents with org charts, budgets, tickets. |
Pure plays on this layer (9)
| Company | Founded | Also on | Influence | Why it is here |
|---|---|---|---|---|
| LangChain | 2022 | — | 6/10 | Agent/chain orchestration — loved and loathed. |
| Scale AI | 2016 | — | 6/10 | Data labeling → RLHF vendor. $870M Meta investment (2025). |
| CrewAI | 2023 | — | 5/10 | Multi-agent framework. Role-playing agent crews. |
| LangGraph | 2024 | — | 5/10 | Stateful graph-based agent orchestration from LangChain. |
| LlamaIndex | 2023 | — | 5/10 | RAG-first data framework. |
| Pinecone | 2019 | — | 5/10 | Managed vector DB pioneer. |
| AutoGen | 2023 | — | 4/10 | Microsoft Research multi-agent conversation framework. |
| Weaviate | 2019 | — | 3/10 | Open-source vector database. |
| XR Blocks (Google) | 2026 | — | 3/10 | WebXR + Gemini framework for AR prototyping. Spatial-AI on-ramp. |
What happened here (21 events)
| Year | Event | Magnitude | What happened |
|---|---|---|---|
| 2016 | TensorFlow + PyTorch | 2 | The framework wars. PyTorch will win research, TF industry (briefly). |
| 2022 | LangChain released | 1 | Agent framework boom begins. (LangChain) |
| 2023 | Vercel AI SDK | 1 | Streaming-first JS/TS framework for AI apps. (Vercel) |
| 2023 | AutoGen + CrewAI arrive | 1 | Multi-agent frameworks proliferate. Agents start talking to each other. (CrewAI) |
| 2024 | Anthropic ships MCP | 2 | Model Context Protocol — USB-C for LLMs. (Anthropic) |
| 2024 | LangGraph released | 1 | Stateful graph agents — the infrastructure layer for long-running workflows. (LangGraph) |
| 2024 | NVIDIA Project GR00T | 1 | Humanoid foundation-model platform + Isaac Sim. NVIDIA becomes the robotics dev cloud. (NVIDIA GR00T · Isaac) |
| 2025 | Scale AI $870M Meta investment | 1 | Data-labeling becomes strategic infrastructure. (Scale AI) |
| 2025 | Paperclip — agent labor mgmt | 1 | MIT-licensed org-chart + budgets + tickets for AI workers. The HR layer for agents. (Paperclip) |
| 2026 | MCP ecosystem 2,000+ servers | 1 | De-facto standard for tool use across vendors. (Anthropic) |
| 2026 | Google XR Blocks v0.12 released | 1 | WebXR + Gemini. Samsung Galaxy XR ships — spatial computing finally ships with AI native. (XR Blocks (Google)) |
| 2026 | SpaceX swallows Cursor for $60B | 2 | Largest startup acquisition ever, four days after SpaceX's own IPO. Every lab's biggest distribution channel now lives inside a rival. (Cursor) |
| 2026 | AgentKernel: The Trust-Native Agentic Operating System | 1 | Architects explicit trust boundaries for autonomous agent runtimes handling privileged actions and untrusted web inputs. |
| 2026 | Agentic Detection of Online Conspiracies | 1 | Multi-agent framework analyzes latent rhetorical structures and coordinated messaging patterns across conversational graphs. |
| 2026 | ZooWork-ShopRanker: An Open, Preference-Aligned E-Commerce Reranker | 1 | Open rerankers trained for general web retrieval transfer imperfectly to e-commerce, where ranking decisions depend not only on topical r... |
| 2026 | IndicBankBench: Evaluating Safety and Reliability of Language Model Assistants in Indian Retail Banking | 1 | Banking assistants must use account-specific information to answer requests and, in many cases, take actions through tools. Evaluating on... |
| 2026 | TRACE: Temporal Audit and Condition-aware Evaluation of Streaming Video Understanding | 1 | Streaming video understanding requires models to interpret evidence as it arrives, yet current evaluations often report task scores witho... |
| 2026 | Evidence-Grounded Auditing of Identification Assumptions in Climate-Policy Causal Evaluations | 1 | Difference-in-differences (DID) studies are widely used to evaluate climate policy, but assessing the evidence supporting their identific... |
| 2026 | How Reproducible Are Evaluation Conclusions? A Self-Audit of LLM-Inferred Prompt Structure | 1 | Evaluations of LLM systems routinely average over small prompt sets and report models as a ranked table. We ask how much confidence such ... |
| 2026 | WideSWE: Can Coding Agents Coordinate Changes Across Repositories? | 1 | Coding-agent evaluation has progressed from resolving individual issues to carrying out long-horizon development, yet task completion is ... |
| 2026 | Groupwise Agentic Grading and Advantage Redistribution for Code Agent RL | 1 | Reinforcement learning (RL) for code agents often uses executable tests to provide binary rewards. With these rewards, Group Relative Pol... |
Claims that rest on this layer
- Layers are value capture, not operations — Companies sit where they capture value, not where they operate. That single rule is what makes the cake a map instead of a directory.
The rest of the stack
- 1 · Silicon — GPUs, TPUs, ASICs
- 2 · Cloud & Compute — Hyperscalers, neoclouds
- 3 · Foundation Models — The frontier labs
- 4 · Tooling & Orchestration — you are here
- 5 · Applications — End-user products
Or read the full index, or open the interactive view filtered to this layer.
The Five Layer AI Cake is published by ConeApex, the trading name of Nirukta Digital XR Technologies Private Limited (CIN U62090TS2025PTC198275), Hyderabad, Telangana, India. The dataset is licensed CC BY 4.0 — cite it freely. See llms.txt for how to attribute it.