For Claude

Run your help desk from a chat.

The odin-api skill teaches Claude how to talk to Odin. Ask it to triage this morning's tickets, look up a customer, or tell you which endpoints stopped checking in — and it works the real API, as you, with exactly the permissions your account already has.

Install it

One line. It drops the skill into ~/.claude/skills/odin-api and touches nothing else.

curl -fsSL https://www.odinhelp.com/skill/install.sh | bash

Pass --dir to install somewhere else, or --force to overwrite an existing copy. The script refuses to clobber an install by default and leaves nothing behind if a download fails.

Or as a Claude Code plugin

Installs the same skill and keeps it updated.

/plugin marketplace add RivetedTech/odin-claude-plugin
/plugin install odin-api@odin

Connect it to your account

Create a token in the Odin dashboard under Profile → Personal Access Tokens, then export it along with the API address for your environment:

export ODIN_API_TOKEN="opat_..."
export ODIN_API_BASE_URL="https://api.odinhelp.com"

# confirm it works
~/.claude/skills/odin-api/scripts/odin preflight

Signed in already? The dashboard's Claude skill button can hand you a download with the token and address filled in for you.

What you can ask for

List, search, create and update tickets
Look up customers, contacts and their history
Check RMM agents and device health across the fleet
Read dashboard and analytics data
Work with invoices and agreements
Audit RentRedi connections and spot the broken ones

About that token

It authenticates as you and carries your live permissions — no more, no less. If a call comes back 403, that is your account lacking the access, not the skill misbehaving.

Treat it like a password: never commit it or paste it into a shared channel. You can revoke it at any time from the same Profile screen, and it expires after a year by default. Point ODIN_API_BASE_URL at your dev environment while you experiment — writes hit real data.

Prefer to read it first?

Nothing here is obfuscated — it is four plain-text files. Read them before you run anything:

Needs bash and curl. jq is optional and only makes the output prettier.