> ## Documentation Index
> Fetch the complete documentation index at: https://docs.weavable.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Contexts

> Create a focused slice of your work data that AI can reason from

Contexts are how Weavable knows what data to reason from — a focused slice of your work data that you query from your AI stack. Different jobs need different data, so you build a context per job: one for renewals, one for support themes, one for engineering activity.

## Create a context

Just describe what you need, the same way you'd prompt any AI. For example:

> *Monitor deal pipeline progress and customer engagement activities using HubSpot and Zendesk support tickets.*

Weavable detects the relevant data sources from your connected apps and walks you through three steps:

<Steps>
  <Step title="Open the context creation page">
    Go to the [context creation page](https://weavable.thatworks.ai/0/context/new).
  </Step>

  <Step title="Describe what you want to create">
    On the **What do you want to create?** screen, type a natural-language description of the context.

    Prefer this to **Or create manually** unless you already know exactly which data sources you want.
  </Step>

  <Step title="Confirm the apps">
    Weavable suggests apps based on your description. Confirm them as-is, swap them out, or pick from the full list.
  </Step>

  <Step title="Confirm the data sources">
    Within each app, Weavable may pre-select specific data sources from your description. Confirm them as-is, swap them out, or pick your own — for example, a particular project, board, channel, repository, or pipeline.

    <Tip>
      Always point to data you actually care about, not everything you have access to. Narrower contexts give faster, more accurate answers.
    </Tip>
  </Step>
</Steps>

Your context is now ready. Primary data sources are set, and Weavable has prefilled instructions for how to use the data.

## Edit a context

Open the context to change its name, instructions, or data sources. Each context has two kinds of data sources:

* **Primary data sources** — the main signal the context is built around. These drive what's in scope when the context is queried.
* **Enrichment data sources** — additional context that gets pulled in to make answers richer (people, related items, etc.).

## Test a context

Use **Open Test Playground** to preview what the context can already answer. It's a good proxy for the kind of depth and detail you'll get when you query the context through your AI stack via [MCP](/mcp/overview).

## Share a context

Share a context with teammates already on Weavable so they get access to the same focused slice of data — without having to connect apps or build their own context.

From the context, open **Share This Context** and choose:

* **Selected people in your organization** — share with specific teammates by email
* **Share with everyone in your organization** — share with everyone on your Weavable workspace

Think of it as giving the right people exactly what they need, without exposing anything they shouldn't see.

<Note>
  Sharing a context is different from [sharing a connection](/product/connections/sharing). A shared context lets others query the same scoped data; a shared connection lets others build their *own* contexts on top of your OAuth credentials.
</Note>

## Connect to your AI

When you're ready, hit **Connect MCP** on the context to get your endpoint and per-client setup instructions for [Claude, Claude Code, ChatGPT, VS Code, and other clients](/mcp/connect).