Skip to main content

​
The problem

Modern AI assistants — Claude, Cursor, ChatGPT — are only as good as the context they can reach. Wire up two or three connectors and an assistant works fine. Connect dozens and it breaks down: the agent doesn’t know where to look, silently skips the source that has the answer, or explores for minutes before it finds anything — and every conversation lands on something different. Critical context is scattered across hundreds of sources — SQL and semantic models, dashboards, tickets, docs, call transcripts, emails — with no map tying them together. The result is slow, inconsistent, and expensive answers.

​
What Modus does

Modus is the connecting tissue between all your data. It builds a live context layer across every source — structured (Snowflake, BigQuery, Looker, dbt) and unstructured (Notion, Teams, Confluence, Drive, email) — without moving or copying your data. When someone asks a question, Modus hands the AI a brief — a real-time generated Markdown document with the relevant context for that specific prompt: a dynamically composed map of exactly which sources, tables, and documents to use to answer the request. Your assistant then does the work through your existing connectors — but instead of dozens of exploratory tool calls, it takes a few. Minutes become seconds, token spend drops, and answers become consistent and accurate.
Modus doesn’t replace your warehouse or semantic models — it uses them, and makes them part of one context layer your whole organization can query with confidence, instead of a pile of MCPs scattered across the company.

​
The context layer

The context layer is the heart of Modus. A few things make it different from wiring connectors straight into an AI assistant:

A map, not a copy

Modus doesn’t move or copy your data. It builds a searchable index and a knowledge graph of where things live and how they connect — across systems — then fetches through your existing connectors on demand.

Built automatically, kept fresh

The context layer is constructed automatically when you connect a source, and updates itself as things change — new tables, new documents — so you’re not stuck maintaining it by hand.

Composed on the fly

For each question, Modus composes a brief — only the sources, tables, and documents relevant to that request — instead of dumping everything at the agent.

Structured + unstructured

It ties your warehouse and semantic models together with docs, tickets, chats, and transcripts, so an answer can span both worlds.

​
Two ways to use Modus

Modus meets you where you already work. Both modes run on the same context layer — pick either, or mix them.

Use your own agent

Keep using Claude, Cursor, ChatGPT, or any MCP client. Connect Modus as a single MCP server and your agent reaches your entire context layer through one connection — no juggling dozens of separate MCPs. Modus does the mapping and hands your agent the right context behind the scenes.

Use Modus

Work directly inside Modus. Chat on Home, tag scopes for specialist help, and run workflows — a Modus-native experience on the same context.
You don’t have to choose: publish a scope and expose it over MCP, so your external agent gets a focused, governed slice of context instead of everything. And because Modus is API-, SDK- (Python & TypeScript), MCP-, and CLI-first, you can automate and build on any of this however you like.

​
How it fits together

Modus is organized around five pages in the left sidebar, each building on the one before it:

Integrations

Connect your data sources (Snowflake, BigQuery, Postgres…) and apps (Slack, Teams, Notion…). This is where your data enters Modus.

Context

The context layer Modus builds automatically from your sources — the metrics, tables, relationships, and documents your AI draws on, kept fresh as things change.

Scopes

Specialist AI configurations. Each scope gets just the slice of context, tools, and access it needs — a focused expert instead of one bloated agent.

Workflows

Triggered, multi-step pipelines that put scopes to work automatically — on a schedule, from a conversation, or via webhook.
And you actually work in one place:

Home

The chat surface where you talk to Modus, tag scopes for specialist help, and run workflows — or connect the same context layer to Claude or Cursor.

​
The flow

Integrations → Context → Scopes → Home (or your assistant)
  1. Connect your data and apps in Integrations.
  2. Modus builds and maintains context from them automatically.
  3. Create scopes — specialist AIs scoped to the right context and tools (optional, but sharper).
  4. Work on Home — or from Claude, Cursor, the API, SDK, or CLI — grounded in your data.
New to Modus? Start with Connecting apps or a data integration, then head to Home to put your data to work.