Getting Started in the Console

The Console at app.openma.dev is the easiest way to use openma — no API calls, no curl. This page walks through what’s where and how to get to “first agent running” without leaving the browser.

Sign in

Email + OTP, or Google OAuth. On first sign-in, openma auto-creates a tenant for you. No invite codes or admin setup.

The pages

The left sidebar groups everything into three areas:

Core

• Dashboard — a one-page view of your agents, recent sessions, and usage.
• Agents — list, create, edit, archive. Click an agent to see its config and version history.
• Sessions — every conversation across every agent, with the live event stream when one is running.
• Environments — sandbox templates. Define which Python / Node / apt packages are pre-installed, set network rules.

Configuration

• Skills — upload a SKILL.md plus any supporting files. The platform mounts them into sandboxes that opt in.
• Vaults — secret stores. Add bearer tokens, OAuth tokens, env vars; bind them to specific hosts so the outbound proxy can inject auth without leaking the secret to the model.
• Memory — semantic memory stores per agent.
• Model Cards — bring your own model: Anthropic, OpenAI, MiniMax, or any compatible provider.
• API Keys — manage personal access tokens for the REST API and CLI.

Integrations

• Linear / GitHub / Slack — each has its own setup page. See Connect Integrations.

Create your first agent in the UI

1. Open Agents → New agent.

2. Fill in:

   • Name — anything. Visible to integrations later.
   • Model — pick from your Model Cards (Anthropic Claude is the default if you have an ANTHROPIC_API_KEY in your tenant settings).
   • System prompt — what the agent is for.
   • Tools — leave default (built-in toolset) for now.

3. Click Create.

4. Open the agent → New session. Type a message. Watch the event stream populate in real time.

Versioning

Every save creates a new agent version. Sessions pin the version they were created with — editing the agent later doesn’t change running or replayed sessions.

Common workflows

Add a Vault so an agent can call a private API

1. Vaults → New vault — pick “Bearer token”, paste it, bind it to the host (e.g. api.example.com).
2. Edit your agent → Tools → add a custom tool whose endpoint is on that host.
3. When the agent calls the tool, the outbound proxy injects Authorization: Bearer ... per request. The model never sees the token.

Give an agent code execution

1. Environments → New environment — pick a base image, add pip:numpy or apt:ffmpeg etc.
2. Edit your agent → Environment → select the new environment.
3. The agent now has bash, read, write, edit, glob, grep running inside a per-session container.

Connect a custom MCP server

1. Edit an agent → MCP servers → paste the URL.
2. The platform handshakes with the server and registers each tool as mcp_<server>_<tool>. They become callable on the next session.

Where to go next

Connect Integrations Wire your agent into Linear, GitHub, or Slack so users interact with it where they already work.

Concepts Refresher on Agent, Session, Environment, Vault, Skill, Tool, Harness, Integration.

REST API Everything the Console does, you can do programmatically.