Notion Connected AI workspace for docs, wikis, projects, and enterprise search (Notion AI, Agents, Q&A).
Notion ships one of the most-watched applied-AI stacks in SaaS: a deliberately model-agnostic layer over OpenAI, Anthropic, Google, and xAI frontier models (users pick the model per agent), on top of a heavily custom in-house agent harness rebuilt four-plus times. They buy best-of-breed infrastructure — turbopuffer for vectors, Braintrust for evals, Fireworks for fine-tuned model serving, Anyscale-managed Ray for embeddings — while keeping orchestration, tools, and retrieval logic in-house. Their subprocessor list is unusually candid, also naming Cerebras, Baseten, Exa, Parallel Web Systems, and MCP security vendor Runlayer.
notion.comSources last verified Jul 27, 2026
Receipts · 3
“The latest—like Claude Sonnet 4 and GPT-5—are already built in” — notion.com, Sep 18, 2025
“Opus 4.6 excels at interpreting what users actually want, producing shareable content on the first try” — claude.com, Jul 2026
“hosted by Notion as well as by organizations such as Anthropic and OpenAI” — notion.com, Jul 2026
Source last verified Jul 26, 2026
Receipts · 2
“The latest—like Claude Sonnet 4 and GPT-5—are already built in” — notion.com, Sep 18, 2025
“Service provider for hosting large language models and embeddings and for abuse prevention” — registora.com, Jul 9, 2026
Source last verified Jul 26, 2026
Receipts · 1
“Service provider for hosting large language models and embeddings” — registora.com, Jul 9, 2026
Source last verified Jul 26, 2026
Fireworks AI Hosts and serves fine-tuned and open-weight models for low-latency features.
Receipts · 2
“By fine-tuning models, we reduced latency from about 2 seconds to 350 milliseconds” — fireworks.ai, Jul 25, 2025
“Service provider for hosting large language models and embeddings” — registora.com, Jul 9, 2026
Source last verified Jul 26, 2026
Anyscale (Ray)
by Anyscale Runs the near-real-time embeddings indexing pipeline on managed Ray.
Receipts · 3
“we set out to migrating our near real-time embeddings pipeline to Ray running on Anyscale” — notion.com, Feb 19, 2026
“Ray lets us run open-source embedding models directly, without being gated by external providers” — notion.com, Feb 19, 2026
“pull a model from Hugging Face and run it ourselves on-prem” — anyscale.com, Jul 2026
Source last verified Jul 26, 2026
In-house agent framework behind Notion Agents with extensive internal tools.
Receipts · 2
“There's now like over a hundred tools. Just for all, all the crazy notion stuff.” — latent.space, Apr 15, 2026
“We have rebuilt our harness three or four times.” — latent.space, Apr 15, 2026
Source last verified Jul 26, 2026
Official hosted MCP server letting external AI clients act on workspaces.
Receipts · 2
“manages sessions and securely stores the API token from the OAuth exchange” — notion.com, Jul 15, 2025
“Official Notion MCP Server” — github.com, Jul 2026
Source last verified Jul 26, 2026
Claude Managed Agents
by Anthropic Embeds Anthropic-hosted coding agents inside Notion via the Managed Agents API.
Receipts · 1
“Managed Agents was great because we just pull in the API and it works within the product.” — claude.com, Jul 2026
Source last verified Jul 27, 2026
turbopuffer Primary vector database for Notion AI search and Q&A.
Receipts · 3
“we committed to migrating our entire multi billion object workload to turbopuffer in late 2024” — notion.com, Feb 19, 2026
“stores it in a vector database (e.g., Turbopuffer)” — notion.com, Jul 2026
“Vector database for storing embeddings” — registora.com, Jul 9, 2026
Source last verified Jul 26, 2026
OpenAI embeddings API
by OpenAI Embeds workspace pages for Q&A retrieval.
Receipts · 1
“we generate an embedding by using an OpenAI zero-retention embeddings API” — notion.com, Jul 2026
Source last verified Jul 26, 2026
Receipts · 1
“Real-time updates via Kafka consumers that process individual page edits as they happen” — notion.com, Feb 19, 2026
Source last verified Jul 26, 2026
Receipts · 1
“Service provider for providing web search results” — registora.com, Jul 9, 2026
Source last verified Jul 26, 2026
Braintrust Eval and LLM-observability platform for regression and frontier-model testing.
Receipts · 1
“I sat down in Braintrust and looked at some of the worst experiences our customers had” — braintrust.dev, Jul 2026
Source last verified Jul 26, 2026
Fine-tunes small open-weight models for search, routing, and function calling.
Receipts · 2
“By fine-tuning models, we reduced latency from about 2 seconds to 350 milliseconds” — fireworks.ai, Jul 25, 2025
“Our, um, fine tuned and open source models are served on GPUs, right?” — latent.space, Apr 15, 2026
Source last verified Jul 26, 2026
Opt-in program feeding customer workspace data into model improvement.
Receipts · 1
“allows for sharing of workspace data to improve underlying models” — notion.com, Jul 2026
Source last verified Jul 26, 2026
Enforces zero-retention and no-training contracts with AI subprocessors.
Receipts · 2
“our LLM providers utilize zero data retention for Enterprise plan workspaces” — notion.com, Jul 2026
“prohibit the use of Customer Data to train their models” — notion.com, Jul 2026
Source last verified Jul 26, 2026
Vercel Sandbox
by Vercel Runs untrusted Notion Workers code in isolated sandboxes.
Receipts · 1
“Under the hood, every Worker runs on Vercel Sandbox .” — vercel.com, Mar 12, 2026
Source last verified Jul 27, 2026
Receipts · 1
“Security for MCP connections” — registora.com, Jul 9, 2026
Source last verified Jul 26, 2026
Claude Code
by Anthropic Fastest-growing agentic coding assistant among Notion's engineers.
Receipts · 1
“Claude Code and Codex Are Outpacing Cursor Among Notion's Engineers” — theinformation.com, Mar 24, 2026
Source last verified Jul 26, 2026
Receipts · 2
“Anyscale lets me use my preferred tools like VSCode or Cursor on my laptop” — anyscale.com, Jul 2026
“Claude Code and Codex Are Outpacing Cursor Among Notion's Engineers” — theinformation.com, Mar 24, 2026
Source last verified Jul 26, 2026
Employees run the company on its own AI agents before public release.
Receipts · 2
“No one uses Notion in their job as much as people that work at Notion.” — latent.space, Apr 15, 2026
“Everyone is using the same instance of notion with like a lot of flags on for these prototypes people build” — latent.space, Apr 15, 2026
Source last verified Jul 26, 2026
