Command-Line Interface

The Querri CLI (querri) is a command-line tool for working with Querri from your terminal. It does everything the web app does — upload files, run analyses, manage dashboards, share resources, manage users — and outputs results in a format that’s easy to pipe into other tools or scripts.

What’s a CLI good for?

A CLI (command-line interface) is a program you run from a terminal by typing commands. You’d use the Querri CLI when you want to:

• Automate Querri — run the same analysis on a schedule, drop a CSV into a project nightly, refresh a dashboard from a cron job.
• Script onboarding — provision users, create access policies, and seed data in one command.
• Stay in the terminal — ask your data a question without switching to the browser.
• Pipe results into other tools — JSON output flows into jq, spreadsheets, dashboards, alerting systems.

If you’d rather embed Querri in a Python application or build a backend that calls Querri’s API, you probably want the Python SDK — same package, importable as a library.

Installation

Install with pip — the [cli] extra pulls in the dependencies the CLI needs (Typer, Rich, Pillow):

pip install "querri[cli]"

Verify the install:

querri --version

If you only need the Python library — not the command-line tool — install the bare package instead:

pip install querri

See the Python SDK page for library usage.

Upgrading

pip install --upgrade "querri[cli]"

First-time setup

Authenticate

The fastest way to log in is browser-based SSO:

querri auth login

This opens a browser, you complete your normal Querri login, and a refresh token is saved to ~/.querri/tokens.json. The CLI auto-refreshes tokens, so you usually only do this once per machine.

Confirm who you’re logged in as:

querri whoami

Use an API key instead (for scripts and CI)

For non-interactive environments — CI jobs, cron, automation servers — use an API key. Create one in the Querri web app under Settings → API Keys, then export:

export QUERRI_API_KEY=qk_your_api_key
export QUERRI_ORG_ID=org_your_org_id

The CLI uses these automatically. The API key takes precedence over a stored login token.

Quick start

Upload a file, create a project, and ask a question:

querri file upload sales.csv
querri project new "Q3 Sales Analysis"
querri project add-source <file_id_from_upload>
querri project chat -m "What are the top 5 products by revenue?"

The CLI keeps an “active” project once you create or select one, so subsequent commands don’t need a project ID. Switch active projects with querri project select.

Scripting

For automation, two flags matter:

• --json — machine-readable JSON output instead of formatted tables.
• --no-interactive — never prompt for input; fail fast if something is missing.

Both are global flags and must come before the subcommand:

querri --json --no-interactive file upload data.csv         # correct
querri file upload data.csv --json                          # WRONG — flag in wrong position

Pipe JSON into jq to extract IDs:

FILE_ID=$(querri --json --no-interactive file upload data.csv | jq -r .id)
PROJECT_ID=$(querri --json --no-interactive project new "Nightly" | jq -r .id)
querri --json --no-interactive --project "$PROJECT_ID" project add-source "$FILE_ID"
querri --json --no-interactive --project "$PROJECT_ID" project chat -m "Summarize the data"

Command reference

The CLI is self-documenting. Run querri --help for the full list, and querri <command> --help for any subcommand.

Authentication

querri auth login                    # browser-based login
querri auth logout                   # clear stored tokens
querri whoami                        # show current user, org, and host

Projects

querri project new "Name"            # create + auto-select
querri project list                  # list projects you have access to
querri project select <name|id>      # set the active project
querri project get [id]              # detail view (defaults to active)
querri project show [id]             # render the analysis step DAG
querri project run [id] --wait       # execute the project's pipeline
querri project run-status [id]       # check execution status
querri project add-source <file_id>  # attach an uploaded file to a project
querri project delete <id>           # delete a project

Files

querri file upload data.csv          # upload a single file
querri file upload "data/*.csv"      # glob upload
querri file list                     # list uploaded files
querri file get <file_id>            # detail + column info
querri file delete <file_id>

Supported formats: CSV, Excel (.xlsx/.xls), JSON, Parquet, and others.

Chat — talk to your data

querri project chat -m "Question"            # send a message in the active project
querri project chat -m "..." --new           # force a new chat session
querri project chat -m "..." --reasoning     # include reasoning traces
querri project chat show                     # show conversation history

To manage chats across projects:

querri chat list [project_id]
querri chat get [project_id] [chat_id]
querri chat new [project_id]
querri chat delete [project_id] [chat_id]

Sources

A source is a connected dataset — either ingested from an uploaded file or synced from a connector.

querri source list                              # list all sources
querri source get <source_id>                   # detail
querri source describe <source_id>              # schema + row count
querri source data <source_id>                  # preview rows (paginated)
querri source query --source-id ID --sql "SELECT * FROM data LIMIT 10"
querri source ask <source_id> "How many rows do we have?"
querri source sync <source_id>                  # trigger a sync
querri source delete <source_id>
querri source connectors                        # list available connector types

In source query, the source is always exposed as a view named data. Always write FROM data, never the source’s display name.

Views

A view is a named SQL query over your sources. Views can be authored either by writing SQL directly or by describing what you want to an AI agent.

querri view new --prompt "monthly revenue by product line"      # AI authors SQL
querri view new --name "Orders" --sql "SELECT * FROM orders"    # write SQL directly
querri view chat <uuid> -m "filter to active customers only"    # iterate via chat
querri view list
querri view get <uuid>
querri view preview <uuid>
querri view run                                                  # materialize all views
querri view delete <uuid>

Dashboards

querri dashboard list
querri dashboard new --name "Sales" --project <id>
querri dashboard refresh <dashboard_id>
querri dashboard delete <dashboard_id>

Sharing

querri share project add <project_id> --user-id <user_id>
querri share project list <project_id>
querri share project remove <project_id> <user_id>

querri share dashboard add <dashboard_id> --user-id <user_id>
querri share source add <source_id> --user-id <user_id>
querri share source org <source_id>             # share with the entire org

Users

querri user list
querri user new --email user@example.com --first-name Alice --last-name Smith
querri user update <user_id> --role admin
querri user delete <user_id>

Access policies (row-level security)

querri policy list
querri policy new --name "APAC Only" --source-ids <id> --row-filters '[{"column":"region","values":["APAC"]}]'
querri policy assign <policy_id> --user-ids <user_id>
querri policy resolve --user-id <user_id> --source-id <source_id>
querri policy delete <policy_id>

API keys

querri key list
querri key new --name "CI" --scopes "admin:projects:read,admin:projects:write"
querri key delete <key_id>

Embedded analytics sessions

For white-label embedding workflows:

querri session new --user-id <user_id> --ttl 3600
querri session refresh --token <token>
querri session revoke --session-id <session_id>

Usage and audit

querri usage org                     # org-wide usage report
querri usage user <user_id>          # per-user usage
querri audit list                    # audit log events
querri audit list --action "project.create"

Environment variables

Variable	Purpose
QUERRI_API_KEY	API key (preferred for scripting; takes precedence over stored login tokens)
QUERRI_ORG_ID	Organization ID
QUERRI_HOST	Server host (default: https://app.querri.com; for local dev: http://localhost)
QUERRI_PROJECT_ID	Active project override (same as --project)
QUERRI_CHAT_ID	Active chat override (same as --chat)

Tips

• Add --help to anything. querri project --help, querri view chat --help, etc.
• --quiet outputs IDs only. Useful inside scripts where you want a single value.
• --verbose prints request details. Helpful when something isn’t working as expected.
• Don’t pass --api-key on the command line. It can land in shell history. Use the QUERRI_API_KEY env var instead.

Talking to your data from an AI assistant

If you’d rather ask Querri questions from inside Claude, ChatGPT, Cursor, or another AI tool — without scripting — use the MCP Server. It exposes the same data through the Model Context Protocol so AI assistants can analyze your data on your behalf.