generated: '2026-09-06'
method: searched
source: https://docs.dremio.com/dremio-cloud/ai-integration/cli/
name: Dremio Developer CLI
binary: dremio
package: dremio-cli
repository: https://github.com/dremio/cli
description: >-
  A first-party command-line interface that wraps the Dremio REST API into commands for queries,
  catalog management, Reflections, engines, users, roles, grants and projects. Dremio documents it
  as an AI-integration surface alongside the MCP server, not just a human tool — it outputs JSON
  by default and ships a `chat` command that drives Dremio's own AI agent from a terminal or a
  pipeline.
requirements:
  python: '>=3.11'
install:
- method: pipx
  command: pipx install dremio-cli
  recommended: true
- method: uv
  command: uv tool install dremio-cli
- method: pip
  command: pip install dremio-cli
  note: >-
    The docs warn that pip into a system Python may fail with externally-managed-environment on
    modern macOS/Linux, and recommend pipx or uv instead.
authentication:
  method: personal-access-token
  required:
  - PAT
  - project id
  precedence: CLI flags > environment variables > config file
  config_file: ~/.config/dremioai/config.yaml
  config_shared_with: the Dremio MCP Server (same file, no duplicate setup)
  environment:
  - DREMIO_TOKEN
  - DREMIO_PROJECT_ID
  - DREMIO_URI
  setup_command: dremio setup
  endpoints:
  - control_plane: US
    uri: https://api.dremio.cloud
  - control_plane: EU
    uri: https://api.eu.dremio.cloud
commands:
- group: setup
  commands:
  - dremio setup
  description: Interactive credential setup — region, PAT, project id discovery, validation, config write.
- group: query
  commands:
  - dremio query run "<sql>"
  - dremio query run --file query.sql
  - cat query.sql | dremio query run -
  description: Run SQL from an argument, a file, or stdin.
- group: search
  commands:
  - dremio search "<term>"
  description: Search the catalog.
- group: schema
  commands:
  - dremio schema describe <path>
  - dremio schema lineage <path>
  description: Describe a dataset's fields and trace its lineage.
- group: job
  commands:
  - dremio job list
  description: List and filter jobs (for example --status FAILED).
- group: reflection
  commands:
  - dremio reflection list <path>
  - dremio reflection refresh <reflection-id>
  description: List Reflections on a dataset and trigger a refresh.
- group: chat
  commands:
  - dremio chat
  - dremio chat --message "<prompt>"
  - dremio chat list
  - dremio chat history <id>
  - dremio chat delete <id>
  description: >-
    Converse with Dremio's AI agent. Interactive mode carries slash commands (/new, /list,
    /continue, /history, /cancel, /delete, /info, /help, /quit); non-interactive mode streams a
    single message to stdout for scripting. --auto-approve approves the agent's tool calls,
    --conversation resumes a session, --model overrides the model, --log-file writes JSON-lines
    events.
global_flags:
- flag: --output
  values: [json, csv, pretty]
  default: json
  note: JSON by default, which is the agent- and pipeline-facing choice.
- flag: --fields
  note: Dot-notation projection, e.g. --fields fields.name,fields.type
- flag: --dry-run
  note: >-
    Validates a destructive command before executing it. The docs demonstrate it on
    `dremio reflection refresh`. This is a real dry-run mode, documented as such.
- flag: --token
- flag: --project-id
- flag: --uri
agent_notes: >-
  The docs publish a worked "autonomous data investigation" sequence (search → schema describe →
  schema lineage → query run → job list → reflection list → reflection refresh --dry-run →
  reflection refresh) as the intended agent flow.