Docs

Quick Start

From an empty organization to a live agent and a connected partner site — the whole path, in order.

Before You Begin

What you will set up, and the one thing you need first.

This page is the day-one path for a brand-new organization. It takes you from the very first account through a running agent and a partner site connected to it, in the order the platform expects. Each step links to the page that covers it in depth — follow them in sequence the first time, and use them as reference afterwards.

The only prerequisite is access to your KeenAgents admin console. Everything below happens there, except the partner steps, which a partner developer does on their own site. Prefer to hand the whole setup to an AI agent? Every step below can be performed by an assistant through the Keen MCP Server — see the Keen Mastering Toolkit.

Set Up Your Organization

Eight steps, in order — each unlocks the next.

  1. Activate the owner account. The very first account is the owner, created through a one-time activation screen while the organization is still empty — you set an email and password and confirm through the emailed link. Once an owner exists that screen closes for good: everyone else joins by invitation, never by self-signup. There is exactly one owner, and it is created here.
  2. Invite your team and set roles. Under Organization, first create the Roles your team needs — a role is a set of permission scopes, each with an Assigned (read) and a Write level. Then create Accounts and give each one a role. Owner, admin and the default role are protected and cannot be deleted. Optionally create an MCP Account (Organization → MCP Accounts): a name, a role and an Enabled switch; you receive its id and a one-time secret to hand to the AI agent that will work for you.
  3. Create a Space. Under Ai, create a Space — the container an agent lives in. Each space carries a deployment token you will use to publish projects into it, and deleting a space cascades to the agents inside it.
  4. Create an Agent. Inside the space, create an Agent: give it a display name and a slug (its stable id) and point it at the flow and node it starts from. It is created Active; the Active switch on its detail page takes it out of service later without deleting it. Its Start Flow / Start Node must match a real flow and start node in the project you'll deploy — don't invent values, or the agent has nothing to run. Optionally define its Json Settings (custom fields your scripts read) and Front Settings (partner-facing configuration) on their own pages.
  5. Build and deploy a project. Build the agent's flows in your project, then deploy it into the space using that space's deployment token. The project is scanned and promoted live in one move; the platform picks it up on its own. This is the Deployment flow under Building Agents.
  6. Open API access. Under Api, create the four pieces a partner integration needs, in this order: a Certificate (the tenant's signing material), a Client (the partner server's identity), a Consumer Contract — on its detail page tick the space from step 3 as Assigned (and as a default if new consumers should get it automatically) and set the five registration/login policy rows — and a Webhook (the partner's browser Origin plus its notification URL). Then create an Api Key and bind all four to it. Only when all four are bound is the key fully bound and usable — and its Origin joins the allow-list. A consumer can only reach agents in spaces its contract grants.
  7. Hand the partner its starting kit. On the Api Key's page click Rebuild Resource Cache (it rebuilds the served copy of everything bound to that key and refuses a half-wired key), then Wake — the partner site receives its certificate material and a notice to read its configuration. Repeat this pair after every later change to anything bound to the key.
  8. Connect your site. Hand the partner its Client id and secret and the Api Key id and secret. The partner's server exchanges its credentials for a token, its site fetches the resources bound to the key, and it opens the live chat against your agent. This is the Connecting a Site walkthrough. If an AI agent will test the agent headlessly instead of a website, it creates its own MCP consumer through the Keen MCP server — no email or one-time pin, just an id and a secret — and mints that consumer's token there.
Owner activation happens once
The activation screen is available only while the organization has no accounts. After the first owner exists it closes permanently, and every later account is created by invitation from inside the admin. Secrets shown along the way — the plaintext API key, a client secret, an MCP account's secret — are revealed exactly once; copy them when you see them.

Where To Go Next

Each step, in depth.

Downloads & Setup

What to install, what the three downloads are for, and how to set up a bot that builds agents.

Accounts

Owner bootstrap, invitations, roles and the account lifecycle.

Roles & Permissions

The permission scopes behind every admin action.

Spaces

The container agents live in, and its deployment token.

Agents

Agent identity, run entry, and the two settings channels.

Deployment

Build, scan and go live with the space token.

Connecting a Site

Wire a partner website end to end, in ordered steps.

Keen MCP Server

Let an AI agent do the console steps for you under an MCP account.

Resource Cache

The served copy partners read — and the Rebuild that refreshes it.

Webhooks

The partner's notification URL and the Wake notice that tells it to re-read.

Previous

Welcome

Next

Downloads & Setup

Keen Agents 2026

Documentation

Release 15