Reltio CLI

The command line interface for Context Intelligence and Unified Data.

Agent-native.Customer first.

The tool AI agents reach for to talk to Reltio.

Your teams already work with AI agents — Claude Cowork, Claude Code, Codex, Copilot CLI — and those agents reach the systems around them through the shell. Reltio CLI meets them there rather than asking anyone to move: one binary added to the setup you already run, with nothing to migrate and no new place to sign in.

$ curl -fsSL https://reltio.tools/install.sh | sh

Or ask your agent:

Install the Reltio CLI from https://reltio.tools/install.sh

macOS on Apple Silicon, v0.10.0 — Windows and Linux come with v1. No account to download, and the same command updates it.

Secure by design.Reltio owns it.

Sign in once. No keys change hands.

reltio login opens a browser, asks for your tenant and works out the rest — whichever sign-in that tenant uses, Reltio's own or your company's SSO. The token lands in a file only your account can read, never in a flag and never in your shell history — and where it travels is settled when you sign in, not when a command runs. reltio get takes a path and refuses a URL, and no environment variable names a host either. The token goes only to Reltio's own hosts under reltio.com, so no prompt can talk an agent into sending it anywhere else. Holding that session is the whole job; everything else goes straight through, byte for byte. Your agent reads your tenant itself rather than this tool's summary of it — and because nothing here models the API, every endpoint works the day Reltio ships it.

zsh — reltio
$ reltio login
# name your tenant in the browser, then sign in
opening https://reltio.tools/login …
waiting for the browser on http://127.0.0.1:38471/callback …
signed in with: browser
tenant: acme
expires_in: 3600s

HTTP-native.Simple and shareable.

Every Reltio API call, ready for stdout.

reltio get sends an authenticated request to any Reltio endpoint and streams JSON to stdout — pipe it to jq, a script, or straight into an agent's context. No flags to repeat once you are logged in.

This is what every skill is built on, and it stays open to you. Reach for it directly when you are exploring, scripting a one-off, or writing a skill of your own — and every call is a line you can paste into a ticket, a runbook or a teammate's terminal.

zsh — reltio
$ reltio get /configuration | jq .sources
[ { "uri": "configuration/sources/CRM", "label": "CRM", "icon": "crm.png", "abbreviation": "CRM" }, … ]

Agent skills.Always better than yesterday.

Your agent already knows Reltio.

Two skills install with the binary. One points an agent at the CLI's own help — every path, every parameter — and the other lets it show you the tenant in your browser. Ask in plain language; it reaches for the CLI itself, and you never open a terminal.

Three more teach the traps no help can — configuration, search, profiles — and install with one variable. A skill is markdown and a few type declarations, so the ones you rely on most can be sharpened by the team that relies on them.

AI agent — acme

How many HCPs are missing an email address?

Skillreltio-search

Callreltio get /entities/_total -q "filter=(equals(type,'configuration/entityTypes/HCP') and missing(attributes.Email))"

1,284. Want them listed, or broken down by source system?

Built in Rust.To be right faster.

One file, no runtime, gone in milliseconds.

Reltio CLI is a single static binary: nothing to install beside it, no Node, no Python, no dependency tree to keep current or to patch. It starts in milliseconds and holds a few megabytes while it runs — which is what you want from a tool an agent calls a hundred times in one session. And the compiler rules out whole classes of crash before a release exists, so what ships either works or never built.