Databricks Feature Store Mcp

naftiko: 1.0.0-alpha2
info:
  title: Databricks Feature Store Mcp
  description: A capability over Azure Databricks feature store, exposing typed model-feature retrieval as MCP for the AI assistant flows John is architecting.
  tags:
  - Naftiko
  - Databricks
  - Feature Store
  - MLOps
  created: '2026-05-01'
  modified: '2026-05-04'
binds:
- namespace: databricks-env
  description: Azure Databricks workspace host and personal access token.
  keys:
    DATABRICKS_HOST: DATABRICKS_HOST
    DATABRICKS_TOKEN: DATABRICKS_TOKEN
capability:
  consumes:
  - namespace: databricks
    type: http
    baseUri: https://{{DATABRICKS_HOST}}
    authentication:
      type: bearer
      token: '{{DATABRICKS_TOKEN}}'
    resources:
    - name: feature-tables
      path: /api/2.0/feature-store/feature-tables/get
      operations:
      - name: get-feature-table
        method: GET
        inputParameters:
        - name: name
          in: query
          description: Three-level feature-table name (catalog.schema.table).
    - name: feature-table-search
      path: /api/2.0/feature-store/feature-tables/search
      operations:
      - name: search-feature-tables
        method: GET
        inputParameters:
        - name: text
          in: query
          description: Free-text search across feature table names and descriptions.
        - name: max_results
          in: query
    - name: features
      path: /api/2.0/feature-store/features/list
      operations:
      - name: list-features
        method: GET
        inputParameters:
        - name: feature_table
          in: query
          description: Three-level feature-table name to list features for.
    - name: online-stores
      path: /api/2.0/feature-store/online-stores/list
      operations:
      - name: list-online-stores
        method: GET
    - name: model-serving-features
      path: /api/2.0/feature-store/feature-tables/get-features
      operations:
      - name: get-features-for-model
        method: POST
        description: Retrieve typed feature values for a set of lookup keys, suitable for online inference.
  exposes:
  - type: rest
    address: 0.0.0.0
    port: 8080
    namespace: databricks-feature-store-mcp-rest
    description: REST API exposing typed feature-store reads from Azure Databricks.
    resources:
    - name: feature-tables
      path: /feature-tables
      operations:
      - method: GET
        name: search-feature-tables
        description: Search feature tables by free-text query.
        inputParameters:
        - name: q
          in: query
          type: string
        call: databricks.search-feature-tables
    - name: feature-table
      path: /feature-tables/{{name}}
      operations:
      - method: GET
        name: get-feature-table
        description: Fetch a feature table's schema, primary keys, online stores, and metadata.
        inputParameters:
        - name: name
          in: path
          type: string
          description: Three-level feature-table name (catalog.schema.table).
        call: databricks.get-feature-table
    - name: model-features
      path: /model-features
      operations:
      - method: POST
        name: get-features-for-model
        description: Retrieve typed feature values for a set of lookup keys.
        call: databricks.get-features-for-model
  - type: mcp
    address: 0.0.0.0
    port: 3010
    namespace: databricks-feature-store-mcp-mcp
    description: MCP server giving AI assistants typed access to the Databricks feature store.
    tools:
    - name: search-feature-tables
      description: Search feature tables in the Databricks feature store by free-text query.
      hints:
        readOnly: true
      inputParameters:
      - name: q
        type: string
        required: true
        description: Search text.
      call: databricks.search-feature-tables
    - name: get-feature-table
      description: Get a feature table's schema, primary keys, source, online stores, and metadata.
      hints:
        readOnly: true
      inputParameters:
      - name: name
        type: string
        required: true
        description: Three-level feature-table name (catalog.schema.table).
      call: databricks.get-feature-table
    - name: list-features
      description: List features (columns and types) defined on a feature table.
      hints:
        readOnly: true
      inputParameters:
      - name: feature_table
        type: string
        required: true
      call: databricks.list-features
    - name: get-features-for-model
      description: Retrieve typed feature values for one or more lookup keys for online inference.
      hints:
        readOnly: true
      call: databricks.get-features-for-model
  - type: skill
    address: 0.0.0.0
    port: 3011
    namespace: databricks-feature-store-mcp-skills
    description: Agent Skill bundle giving AI assistants typed feature-store access.
    skills:
    - name: databricks-feature-store-mcp
      description: Typed model-feature retrieval from the Databricks feature store for AI assistant flows.
      location: file:///opt/naftiko/skills/databricks-feature-store-mcp
      allowed-tools: search-feature-tables,get-feature-table,list-features,get-features-for-model
      argument-hint: 'Use when an agent needs to look up feature definitions or fetch typed feature values for inference.'
      tools:
      - name: search-feature-tables
        description: Search feature tables.
        from:
          sourceNamespace: databricks-feature-store-mcp-mcp
          action: search-feature-tables
      - name: get-feature-table
        description: Get a feature table's schema and metadata.
        from:
          sourceNamespace: databricks-feature-store-mcp-mcp
          action: get-feature-table
      - name: list-features
        description: List features on a feature table.
        from:
          sourceNamespace: databricks-feature-store-mcp-mcp
          action: list-features
      - name: get-features-for-model
        description: Retrieve typed feature values for inference.
        from:
          sourceNamespace: databricks-feature-store-mcp-mcp
          action: get-features-for-model