The Semogram plugins
Connect systems and extend Semogram with internal and custom plugin capabilities.
Plugins provide the capabilities Semogram uses to connect to systems, transform records, materialize facts and resolve credentials. One plugin can bundle several capabilities and share one installation configuration.
A plugin defines the implementation and contracts. An installation configures a version in your workspace. A data endpoint names a concrete table, collection, object path or fact-store target. A pipeline node uses an installed capability to perform work.
Setup guides
Use the internal plugin guides for Postgres, MongoDB, S3, News/Web, HTTP Dataset, Iceberg, Apache Jena, Environment Variables and Transforms. Each covers installation, a first bounded operation, connection management and FAQ.
Custom creation
Create a custom plugin when an existing capability does not fit. Define its behavior, generate a canonical package, inspect sandbox validation, publish it into the workspace catalog, install it and verify a real bounded operation.
Reference
- Capabilities and contracts explains installation configuration, selectors and runtime interfaces.
- Transforms reference covers all 20 bundled operations.
- Package format describes
craven.plugin.yaml, runtime exports and bundled skills. - Installation lifecycle covers versions, revisions and removal.
- Configuration and troubleshooting covers checks, credentials and failed runs.
Plugins and MCP
Semogram’s MCP server lets an outside assistant use platform tools. Some tools require installed plugin capabilities; their reference pages mark those requirements. An external MCP-server capability supplies tools to Semogram and is configured separately from an assistant connecting to Semogram.
Installation does not start ingestion, configure every target or grant new source permissions. Verify a bounded operation before scheduling recurring work.