datris-mcp-server

Datris.ai is an open-source agent-native data platform focused on ingestion and data quality. It sits beside an existing warehouse or lake rather than replacing them, and it uses AI agents over MCP alongside developers and analysts. The site says a first pipeline can be set up in about 60 seconds.

Its built-in Assistant Agent is part of the platform UI and handles pipeline setup from a natural-language request. In the example shown, it asks scoping questions before building, chooses a source and destination, generates the fetcher, requests credentials through a secure form, runs the job, confirms that rows landed, and then lets the user query the result. The feature list also includes plain-English validation through aiRule, natural-language row transformations, auto-configuration from uploaded CSV, JSON, or XML files, instant profiling with quality issues and suggested validation rules, and AI error explanations in plain language. The architecture section says Datris handles data acquisition, validation, normalization, storage, and observability.

The platform supports ingestion from MinIO for real-time uploads, Kafka for streaming, and taps for APIs, databases, files, documents, or custom scripts. It also supports RAG workflows with document extraction, chunking, embedding, and upserting into vector databases, and names Qdrant, Weaviate, Milvus, Chroma, and pgvector. On the MCP side, agents can register pipelines, upload files, trigger jobs, profile data, run searches, generate schemas, monitor jobs, and retrieve results. The page also names MCP stdio and SSE, and shows integrations with Claude, Cursor, and OpenClaw.

Datris is fully open source and can be self-hosted. The self-hosted section lists MinIO, MongoDB, ActiveMQ, HashiCorp Vault, Apache Kafka, and Apache Spark as part of its infrastructure, and the page shows GitHub and a Docker-based clone, configure, launch flow.