Documentation Index
Fetch the complete documentation index at: https://docs.3ngram.ai/llms.txt
Use this file to discover all available pages before exploring further.
What are content sources?
Content sources let you connect external repositories (GitHub repos, Google Docs, Confluence spaces) so their content is indexed and searchable by your AI alongside your memories. Your AI gets full context without you having to paste anything in.Connecting a source
Sources are added via the web dashboard at app.3ngram.ai:- Open Sources (the consolidated integrations + content-sources page)
- Connect your GitHub, Google, or Atlassian account, then select the repos or docs to index
- The system automatically chunks and embeds the content for vector search
Checking your sources
Use the MCP tool to see what you have connected:“3ngram: List my content sources”This shows each source’s repo name, branch, sync status, and last indexed time.
Triggering a re-sync
If you want fresh content indexed immediately:“3ngram: Sync source ”This queues the source for re-indexing on the next cron cycle. You can find the
source_id from the list command above.
Searching indexed content
Content from your sources is automatically included insearch results. Use the search tool as normal:
“3ngram: Search for ‘authentication middleware’”Results include both memories and indexed content from your connected sources. The AI uses vector similarity to find the most relevant chunks across everything you have connected.
Supported providers
| Provider | Content Indexed |
|---|---|
| GitHub | Repository files (code, docs, READMEs) |
| Google Docs | Document content |
| Confluence | Page content |
How indexing works
- Content is split into semantic chunks optimized for retrieval
- Each chunk gets a vector embedding for similarity search
- A background cron job keeps content fresh by re-syncing on a schedule
- Per-plan limits apply to the number of connected sources
Content sources require a paid subscription. Connecting and removing sources happens in the web dashboard at /sources. The MCP tools listed above let your AI inspect connected sources (
list_sources) and queue a re-index for one of them (sync_source — which writes to the source’s sync state and is gated by always_ask by default for managed agents).