Build with the API

Rightbrain Agent Skill

Let coding agents build and operate Rightbrain through the live API.

The Rightbrain Agent Skill is the official Agent Skills package for building and operating Tasks, agents, knowledge, tools, and automations from a compatible coding agent. It is also distributed as a Claude Code plugin.

Why use the skill

The skill gives an agent an execution workflow rather than a static copy of the API:

  • It resolves the current operation contract from Docs MCP, then llms.txt, then OpenAPI, stopping when the operation is clear.
  • Its rb-api.mjs wrapper handles Organization and Project scoping, pagination, SSE streams, uploads, file output, timeouts, and redacted dry runs.
  • Its instructions tell the host agent to verify the selected production host, Organization, Project, target, and intended effect before a write.
  • Its instructions require confirmation before deletion, public sharing, credential or permission changes, automation enablement, external writes, and potentially expensive runs.
  • Structured responses printed to stdout redact known secret fields. Use --secret-output for a confirmed one-time-secret handoff to a new protected file.

This is useful when an agent needs to carry out a multi-step platform task without embedding an API snapshot or assembling authentication, pagination, upload, and safety behavior separately for every operation.

Relationship to the CLI

The Rightbrain CLI owns interactive login and session refresh. The Agent Skill uses that session by default and adds live contract discovery, API execution, response handling, and safety checks.

Headless environments can use a secret-managed API key or OAuth client credentials instead. Authentication precedence is:

  1. RB_API_KEY with RB_ORG_ID and RB_PROJECT_ID
  2. RB_CLIENT_ID with RB_CLIENT_SECRET
  3. An authenticated Rightbrain CLI session

The wrapper requires Node.js 18 or newer. CLI session authentication requires Node.js 22 or newer and Rightbrain CLI 0.3.0 or newer.

Install

Claude Code can install the repository as a plugin marketplace:

$claude plugin marketplace add \
>  https://github.com/RightbrainAI/rightbrain-agent-skill.git
$claude plugin install rightbrain@rightbrain-agent-skill

For Cursor, Codex, and other compatible Agent Skills clients, copy plugins/rightbrain/skills/rightbrain from the repository into .agents/skills/rightbrain for one project or ~/.agents/skills/rightbrain for a personal installation.

Install and authenticate the CLI before using session authentication:

$npm install --global rightbrain@latest
$rightbrain login --non-interactive

The public skill instructs the host agent to operate against https://app.rightbrain.ai/api/v1 and stop if whoami reports another host. The wrapper does not enforce the production host or confirmation workflow on its own. Keep API keys and OAuth client secrets in a secret manager, never in a prompt or command argument. The --output option writes response bodies verbatim, so protect and inspect those files separately.

Use and verify

Ask the client to use the Rightbrain skill for a Rightbrain operation. Claude Code users can also run /rightbrain:rightbrain.

The skill’s operating loop begins with whoami. Before a write, confirm that it reports the intended Organization, Project, and production API URL. The host agent then reads the current operation contract, shows the proposed effect when required, executes through the wrapper, and reports the response.

Rightbrain CLI

Configure environments, authenticate, and manage Tasks directly.

Authentication

Select credentials for direct and headless API access.

API Reference

Browse the operations the skill resolves at execution time.

GitHub repository

Read installation, security, release, and contribution details.