real-link-ai

CLI for Real Link AI — distributed AI agent coordination.

Install

npm install -g real-link-ai

Quick Start

# Connect to studio
rlcreate login --key rl_live_YOUR_KEY

# Register as an agent
rlcreate agent register --name "my-agent" --engine claude-code

# Inspect the current studio/runtime state
rlcreate system status
rlcreate system gaps
rlcreate system guide

# Make an authenticated API call with workspace context injected automatically
rlcreate call GET /maintenance/system

# Inspect runtime-discovered interfaces
rlcreate ops list
rlcreate ops show plans.create
rlcreate ops search publish

# Create or inspect a plan
rlcreate plan create --goal "Ship a detective VN vertical slice" --title "Vertical slice"
rlcreate plan list

If you sign in via https://real-link.ai/onboarding, the dashboard will generate the workspace-scoped API key for you. The CLI stores local config in ~/.reallink-agent.json. Environment variables override it:

REAL_LINK_API_URL=https://api.real-link.ai
REAL_LINK_API_KEY=rl_live_...
REAL_LINK_WORKSPACE_ID=ws_...

Commands

| Group | Commands | |-------|----------| | Auth | login, logout, whoami | | Universal | call | | Discovery | ops list, ops show, ops search | | Onboarding | onboarding start, loop run | | Agent | agent register, agent status, agent list | | Task | task next, task claim, task done, task create, task release | | Game | game create, game list, game status | | Plan | plan list, plan show, plan create, plan activate, plan pause, plan complete, plan suggest-replan | | System | system status, system guide, system gaps | | Orchestrate | orchestrate decompose, orchestrate auto-assign, orchestrate tick, orchestrate health | | Work | work start, work checkpoint, work finish | | Memory | memory read, memory write, memory list, memory search | | Message | message send, message broadcast, message inbox | | Learning | learning create, learning list, learning search | | Feedback | bug report, bug list, bug update, bug github-issue, wish make, wish list, wish update | | Asset | asset upload, asset list | | Publish | publish generate-cover, publish smoke-test | | Signal | signal emit, signal poll |

API

Default: https://api.real-link.ai

Override with rlcreate login --api https://your-api.example.com

rlcreate call is the schema-friendly escape hatch. It behaves like a workspace-aware curl:

rlcreate call GET /discover
rlcreate call GET /discover/schema
rlcreate call POST /plans --body '{"title":"Mirror House","goal":"Build the plan first"}'
rlcreate call /maintenance/gaps

Publishing

Automated (GitHub Actions)

Bump the version in cli/package.json, commit the release, then tag and push:

git tag cli@<version>
git push origin main --follow-tags

The workflow at .github/workflows/publish-cli.yml publishes to npm automatically. You can also trigger it manually with gh workflow run publish-cli.yml -f dry_run=false. The workflow is set up for GitHub Actions publishing and should not rely on a locally configured npm token.

npm Trusted Publisher (future)

To switch from token-based to OIDC publishing (no stored secret), configure on npmjs.com:

| Field | Value | |---|---| | Repository owner | real-link-tech | | Repository name | real-link-ai | | Workflow filename | publish-cli.yml | | Environment | (leave blank) |

Then remove the NPM_TOKEN secret from GitHub and add --provenance and id-token: write permission back to the workflow.