Answer Engine integrations start with source types the product can ingest today: Firecrawl web pages, Jira and Confluence, CSV data, and documents. The page is structured as repeatable source records for future programmatic pages, and Slack is shown only as roadmap work rather than a verified connector.
Each source is represented as a small record with slug, status, ingest
mode, and content type. That makes later pages like
/integrations/firecrawl-web straightforward to generate.
Use web page and domain ingestion when the memory source is a site, docs section, help center, or public knowledge base.
Import tickets, specs, and workspace pages so agent answers can cite project decisions and source documents.
CSV imports support repeatable operational datasets, customer lists, export files, and other tabular source material.
Documents become inspectable content items with source metadata, extracted text, summaries, and downstream artifacts.
Slack is intentionally roadmap-labeled here so the page does not imply a shipped connector before it is code-backed.
Integrations feed the memory layer. Agent clients call the memory layer through MCP, using the same install command shown across the marketing site.
Source records preserve metadata and lineage so recall can remain inspectable after the content is indexed.
npx answer-engine-mcp A source is not presented as verified until it is backed by the product. Roadmap connectors can appear only with an explicit Roadmap chip and copy that says the work is planned.
Connect a source, call memory through MCP, and inspect the source trail before expanding to additional integrations.