Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.opper.ai/llms.txt

Use this file to discover all available pages before exploring further.

The Opper CLI is the official opper command for the terminal. Beyond direct platform access — calling functions, managing indexes and models, inspecting traces, tracking usage, generating images — it can also launch coding agents (Claude Code, Claude Desktop, OpenCode, Codex, Hermes, OpenClaw, pi) with their inference transparently routed through Opper.

​
Install

npm i -g @opperai/cli
Or run any command on demand without installing: 
npx @opperai/cli <command>
Requires Node.js ≥ 20.12.

​
Quick start

opper login              # OAuth device flow
opper whoami             # confirm the active account
opper launch claude      # launch Claude Code through Opper

​
What you can do

Command groupWhat it does
login / logout / whoamiOAuth device flow and account inspection.
askAsk a question grounded on the installed Opper skills. Useful for “how do I…” questions about the platform, SDKs, or the CLI itself.
callRun an Opper function from the terminal, with --model and --stream.
launch <agent> / agents list / agents uninstall <agent>Launch coding agents through Opper — Claude Code, Claude Desktop, OpenCode, Codex, Hermes, OpenClaw, pi. agents uninstall removes the Opper integration without uninstalling the agent itself.
editorsWire AI code editors (OpenCode) to use Opper as a provider.
skillsInstall, update, list, or uninstall bundled Opper skills.
functions, models, indexes, traces, usageManage every Opper platform resource directly.
image generateGenerate images from a prompt.
configManage stored API keys and slots.
For full flags on any subcommand: opper <command> --help. Or run opper with no arguments for an interactive menu.

​
Launch coding agents through Opper

opper launch <agent> is the fastest way to route a coding agent through Opper — every model call from the agent goes through your Opper account, with the trace, cost, and model routing benefits that come with it.
AgentSlugHow Opper plugs in
Claude CodeclaudeANTHROPIC_BASE_URL / ANTHROPIC_AUTH_TOKEN env vars
Claude Desktopclaude-desktopwrites a third-party-inference (deploymentMode: "3p") profile into ~/Library/Application Support/Claude-3p/ (macOS) / %LOCALAPPDATA%\Claude-3p\ (Windows); quits and reopens the GUI app to apply
OpenCodeopencodeprovider block in ~/.config/opencode/opencode.json
Codexcodex[model_providers.opper] block in ~/.codex/config.toml
Hermeshermesisolated HERMES_HOME=~/.opper/hermes-home/ so your real ~/.hermes/ is never touched
OpenClawopenclawopper provider entry in ~/.openclaw/agents/main/agent/models.json; defaults to running gateway start (background daemon)
pipiopper provider entry in ~/.pi/agent/models.json
Anything after the agent name is forwarded to the agent’s CLI verbatim (Claude Desktop is GUI-only and does not accept extra arguments): 
opper launch claude -- --print "Hello"
To remove an agent’s Opper integration without uninstalling the agent itself: 
opper agents uninstall claude-desktop   # works for any registered adapter

​
Ask the CLI a question

opper ask "<question>" runs an Opper agent grounded on the locally-installed Opper skills. The answer streams to your terminal, then prints token and request counts. 
opper skills install                                          # one-time
opper ask "how do I create an index?"
opper ask --model anthropic/claude-opus-4-7 "compare the v2 and v3 APIs"
ask is the fastest way to get unstuck without leaving the terminal — the model has the same skill packs available as the agents you launch through opper launch.

​
Authentication

Auth state lives in ~/.opper/config.json as a list of slots, each holding an API key and base URL. Use --key <slot> on any command to pick which slot to use. Resolution at request time: OPPER_API_KEY env var > slot named by --key (or default). 
opper config add staging <api-key> --base-url https://...
opper call --key staging myfunction "instructions" "input"

​
Source and reference

View on GitHub The repository README is the source of truth for releases, supported agents, and configuration details. For exact arguments and flags, opper <command> --help is always authoritative.