> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://api.docs.gooclaim.com/llms.txt.
> For full documentation content, see https://api.docs.gooclaim.com/llms-full.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://api.docs.gooclaim.com/_mcp/server.

# Knowledge Sources Quickstart

The **Knowledge Layer** is Gooclaim's grounded answer engine. It indexes your
policy documents, claim handling manuals, FAQs, and reference data so
Gooclaim's agents can answer questions like "why was this claim rejected?"
with the exact, cited reason — never made up, always traceable.

This quickstart walks through connecting your first source and verifying that
content is queryable from your workflows.

## What you'll connect

Knowledge sources are anywhere your team already stores claim-related reference
material:

SharePoint, Google Drive, Confluence, Notion — for policy PDFs, SOPs,
rate cards, and exclusions lists.

Zendesk, Intercom, internal wikis — for FAQ-style content claimants
typically ask about.

S3, Azure Blob, GCS — for bulk archives of historical claim memos,
medical guidelines, or partner-specific overrides.

Anything reachable over HTTPS with an API key or OAuth — for live
reference data that changes daily.

## Connect a source in 4 steps

From your Portal, open **Knowledge → Sources**. You'll see the catalogue
of supported connectors grouped by category.

Click the connector you want (e.g. SharePoint). You'll be redirected to
that platform's OAuth flow. Sign in with an account that has read access
to the documents you want indexed.

For API-key connectors, paste the key when prompted. Gooclaim stores it
encrypted via AWS Secrets Manager — never in the application database.

Select specific folders, sites, or collections to include. We recommend
starting narrow — a single policy folder — and expanding once you've
verified the agent's answers.

Files are filtered by type at ingest. PDFs, DOCX, MD, HTML, and TXT are
indexed by default. Images and scanned PDFs are passed through OCR.

Click **Sync now**. The first sync pulls every selected document, runs
each through extraction → chunking → embedding → indexing. Progress is
visible per-document in the Sources screen.

Subsequent syncs are incremental — only changed documents re-process.
Schedule recurring syncs (hourly, daily, weekly) under **Source settings**.

## Verify a query

Once the first sync completes, open **Knowledge → Library** and use the
test-query box at the top. Type a question your indexed documents should be
able to answer — for example, "what is the network hospital limit for plan
B?".

You should get back a grounded answer with the exact source document and
section cited. If the answer is wrong or empty, see [troubleshooting](#troubleshooting).

## How this connects to claim workflows

Every reply Gooclaim sends to a claimant is built on three layers:

1. **Truth Layer** — live status from your CMS or claim system
2. **Knowledge Layer** — what you just connected, for context and reasoning
3. **Policy Gate** — final compliance check before anything leaves the platform

When a claimant asks "why was my claim rejected?", the workflow:

1. Pulls the rejection reason from the CMS (Truth Layer)
2. Looks up the policy clause that drives that reason (Knowledge Layer)
3. Formats the answer through an approved template
4. Runs the Policy Gate to confirm no PHI leakage or non-compliant phrasing
5. Sends the templated reply over the channel the claimant used

No free-text generation reaches end users. Knowledge sources only feed the
reasoning step — the surface is always templated.

## Troubleshooting

The most common cause is OAuth scope. Re-run the OAuth flow and confirm
read access to the resource you selected. If the platform requires admin
consent, ask your IT admin to approve.

Check the document was actually indexed — go to **Library**, search for
its filename. If missing, look at the source's last sync report for skip
reasons (unsupported format, encrypted, oversized).

Likely indexing quality. The Knowledge Layer chunks documents semantically,
but heavily-tabular content (rate cards, exclusions matrices) sometimes
needs to be re-uploaded as structured CSV/JSON. Contact [support@gooclaim.com](mailto:support@gooclaim.com)
for help reshaping the source.

## Next steps

Tune query expansion, hybrid search, and metadata filtering for your data.

See how Knowledge sources plug into the three pilot workflows.