Connect a source and choose what gets indexed

Connect a provider, catalog its content, and index only the parts that you need.

When you connect a provider, Orcha does not index all of its content. Orcha first catalogs the source. Then Orcha indexes only the content that you select. This guide shows the procedure in the app. For the model that the procedure uses, read Connected sources.

Each provider also has a guide that shows how to add, use, and remove it: GitHub, Google Workspace, Notion, Zoom, Linear, Asana, Confluence and Jira, Microsoft 365, Dropbox, and Granola.

You must have the workspace admin role.

1. Connect the provider

Open Connectors.
Select a provider card.

Most providers open an OAuth consent screen. GitHub installs a read-only GitHub App on the repositories that you select. Granola uses an API key. If a card shows that setup is incomplete, an operator must configure credentials for your environment.

You can connect more than one account for each provider. Each account has its own tab under Manage.

2. Let the catalog build

The first sync reads the provider and records the objects that it contains. The catalog records names, paths, sizes, and revisions. The first sync does not download the object bodies.

This step has a low cost. The provider's daily call budget sets its limit. The Manage view shows this budget as the calls used today.

3. Set indexing rules

Each resource has a policy. A resource is, for example, a repository, a drive, a Notion workspace, or a Linear team. A path in the resource can have a rule that overrides the resource policy.

Policy	Use it when
Index essentials	You want the documentation and guidance files, but not the full tree
Index all supported content	You want to search the full subtree
Index names only	You want to find and live-fetch the content, but you do not want Orcha to store it
Don't include	The subtree has no useful content, or you must not share it

The deepest rule that matches a path has priority. For example, a repository can use essentials at the root, full content under docs/, and no content under vendor/.

When you select a policy, Orcha shows a preview of the result. The preview shows the number of objects to index, the number of bytes, and an approximate number of provider API calls. The rules apply when you save them.

Each provider has a different default:

GitHub and Granola index essentials.
Drive, OneDrive, Dropbox, Notion, Linear, Jira, and Zoom catalog first. They index nothing until you add a rule.
Gmail is off until you add its OAuth scope and an explicit rule.

4. Check the result

Each object in the catalog shows its state:

Indexed: Orcha stores the body, and you can search it.
Queued: You selected the object. The next sync fetches it.
Catalog only: Orcha stores only metadata. balanced and fresh searches can still get the object.
Skipped: Orcha cannot index this type. The state shows the reason.

To apply changes immediately, click Sync now. If you do not, the changes apply at the next scheduled sync.

All plans sync each hour. Paid plans also accept provider webhooks from GitHub, Dropbox, and Linear. These webhooks update the index between the scheduled syncs.

5. Query it

search_context returns indexed source documents in the same way as other content. Each result cites the exact upstream version. You can also get catalog-only objects without an index:

{
  "query": "how the payments webhook retries",
  "mode": "balanced",
  "sourceIds": ["4f6c2d8a-1b7e-4a52-9c31-0d5f8e2a7b64"]
}

The balanced mode fetches live a small number of catalog entries that match clearly. The fresh mode also sends a live query to providers that support it, for example issue trackers.

To query one provider directly, use query_source. To get the full body of one document, use fetch_source_document. To get source IDs, use get_workspace_overview.

When to index rather than catalog

Index the content that your agents read frequently. Examples are documentation, runbooks, specifications, and meeting summaries. Keep all other content in the catalog only.

Indexed content gives faster results. It uses no provider calls at query time. You can search it when the provider is not available. Catalog-only content has no storage cost, and a query can still get it when necessary.