Connecting external systems

bRRAIn connects to the systems you already run through governed gateways, so the organization's identity, permissions, policy and audit apply on the way in. Pick the route that matches the job:

| You want to | Use | | --- | --- | | Query live rows in a database, warehouse, object store or SaaS tool without copying them | Data Pipe: 31 source connectors, including PostgreSQL, MySQL, SQL Server, Oracle, Snowflake, BigQuery, MongoDB, Elasticsearch, S3, Salesforce, HubSpot, Google Sheets and Kafka. Also available as datapipe_* MCP tools. | | Give AI assistants and agents governed access to the vault and other tools | MCP Gateway and the Exchange MCP catalog of MCP servers | | Use a model the organization has approved | LLM Registry, with the Handler as default | | Bring documents (PDF, Office files) into the knowledge base | Document ingestion through the console, Nexus or Ingestion().Commit | | React to vendor events | Integration webhooks: Integrations().Subscribe from an extension | | Anything custom: a queue consumer, a scheduled sync, a new interface | Write an extension with the Platform SDK |

Writing an extension for a custom source

When no connector fits, an extension reads the source and commits what belongs in the knowledge base through Ingestion().Commit, so the organization's size rule and security gate run on every item:

1. Read source credentials at call time with OperatorEnv().Get or OAuth().ResolveBearer; never store them (Examples).
2. Keep your cursor or offset in extensions/<slug>/state.json with Vault().Write.
3. Commit each item with a stable RecordID and your extension's SourceSlug (Vault and ingestion).
4. Tag what you write with POPE tags (Vault paths and POPE tags).
5. Record the run in the audit chain (Governance).

The bRRAIn Certified Integration Engineer course covers multi-system integration architecture; the SDK Developer course covers building the extension.