Senso CLI

This page covers installing the CLI, signing in and the commands you use most. Every command and flag is in the full CLI reference.

Installation

Requires Node.js 18 or later. Runs on Linux, macOS and Windows.

Install

bash

npm install -g @senso-ai/cli

Sign in

bash

senso login

senso login opens an approval page in your browser and prints a short code. Sign in, type the code and approve. You must be an admin to approve. The CLI then stores your credentials locally, so every command after that is signed in. The code works for five minutes. Credentials last 7 days.

Running it again does nothing if your key still works, so an agent can safely run it at the start of every session. Inside an agent, which cannot wait on a command, it prints the link and code and exits. The agent then runs senso login --complete, which waits for your approval and stores your credentials.

Flags

--complete

Finish a login started without a terminal: wait for the browser approval, then store your credentials.

--api-key <key>

Store a key you already have instead of approving in a browser.

--interactive

Paste an existing API key at a prompt instead. Needs a terminal.

--device-name <name>

How this device is labeled on the approval page.

--no-browser

Print the approval link without opening it.

senso logout — revokes the key login created, then forgets it. Run logout and then login to switch organizations or renew early.
SENSO_API_KEY — set it in the environment to use a key without storing it. It overrides the stored key, and senso whoami shows which one was used.

Global flags

Available on every command. --output json prints structured output for scripts and agents, and --quiet suppresses banners. The agent skills set both by default.

--api-key <key> — use a specific key for one command
--output <json|table|plain> — default is plain
--quiet — hide banners and decoration
--no-update-check — skip the version check

Step 1 — Build the Context Layer

Get your approved sources in and query them. What this step is for →

senso ingest upload <files>

Ingest up to 10 files at once.

senso kb upload <file> --folder-id <id>

Ingest into a specific folder.

senso kb my-files

List your sources and folders.

senso kb find --query "refund"

Find a source by name.

senso kb create-folder --name "Policies"

Create a folder.

senso brand-kit set --data '{…}'

Set voice, tone and writing rules.

senso content-types create --data '{…}'

Add an output template.

senso search "What is our refund policy?"

Grounded answer with sources.

senso search context "mortgage rates"

Raw chunks, no generated answer.

Step 2 — Evaluate & Remediate

Track the questions that matter and fix what they expose. What this step is for →

senso prompts

Manage the questions that drive generation.

senso questions

Manage the questions monitored across models.

senso ingest reprocess <node-id> <file>

Replace a source on an existing document.

senso content get <content-id>

Read one compiled document.

Step 3 — Generate & Verify

Draft against a gap, then check every claim in it. What this step is for →

senso generate sample --prompt-id <id> --content-type-id <id>

Draft one piece.

senso generate run

Trigger a full generation run.

senso content verification

See what is waiting for review.

senso engine draft --data '{…}'

Save a draft instead of publishing.

Step 4 — Publish & Observe

senso engine publish --data '{…}'

Publish verified content.

That is the only command in this step. Measurement lives on the REST API only, so read it with /org/analytics rather than looking for a senso analytics that does not exist.

Account and setup

senso whoami

Check which organization your key resolves to.

senso org get

Organization details.

senso credits balance

What you have left.

senso api-keys list

List keys.

senso skills install quickstart verification-loop-setup verification-loop shared-context-setup shared-context context-layer gap-report generate-verify publish --global

Install the Senso agent skills.

senso update

Update the CLI.

Run senso --help or senso <command> --help for full options.

Limitations

Ingest markdown by file. senso ingest upload rejects .md — send markdown as text through the MCP server instead.
Create an organization. senso org can get, update and set-runs, and nothing more.
Publish on a brand-new organization. engine publish needs a tracked question id, and a fresh org has none.
Report analytics. That is REST only.

Supported upload formats

PDF, TXT, CSV, DOC, DOCX, XLS, XLSX, PPT, PPTX, HTML, JSON and XML. Markdown and plain text go in as text, not as a file upload.

Without a terminal

If your assistant cannot run commands — claude.ai, ChatGPT, Claude Desktop — connect the MCP server instead. It covers step 1 entirely, ingests markdown directly, and needs no terminal. Generation and publishing stay here on the CLI.

CLI Reference Connect Your Agent The Playbook API Reference