Skip to main content

Pre-built Agents

The Agents catalog sits under the Hub · Catalog eyebrow and is described as "Drop-in agents for common enterprise workflows — one-click import into Studio." Each agent is a fully configured, production-ready AI assistant packaged for a specific enterprise use case. Teams browse the catalogue here, review the agent's capabilities and documentation, and then install it from the Marketplace in the Runtime that will host it.

Catalog Statistics​

Three stat tiles above the card grid summarise the current catalog state:

  • Published agents — Total count of agent entries in the catalog.
  • Total capabilities — Sum of all declared capabilities across all published agents.
  • Categories — Count of distinct categories in use across the catalog.

Search and Filter​

A search bar and a category filter sit above the card grid:

  • Search the catalog — Filters cards in real time by agent name or tagline. The input matches partial strings case-insensitively.
  • Category filter — A set of pill buttons, one per distinct category present in the catalog. Selecting a category narrows the grid to entries in that category only. Selecting the same category again returns all entries.

Catalog Card Grid​

Each agent is represented by a card in a responsive grid (one column on mobile, two on tablet, three on wide screens). The card surface shows:

  • Category eyebrow — The agent's category label, rendered in the category accent color.
  • Agent name — The primary identifier for the agent.
  • Tagline — A one-line description of the agent's purpose.
  • Invocation mode badges — One badge per supported channel. Possible values: Chat, API, A2A, Copilot, Teams, Studio.
  • Capability count — Displayed at the bottom of the card alongside the version string in the form N capabilities · vX.Y.
  • View link — Opens the agent detail page.

Agent Detail Page​

Clicking any card navigates to the agent detail page. The header shows the agent name, category eyebrow, tagline, version, capability count, and the agent's invocation mode badges. A Back to catalog link returns to the grid.

Detail Tabs​

Three tabs organize the agent's detail content:

  • Overview — Renders the agent's full description as formatted markdown.
  • Capabilities — Lists all capabilities declared by the agent. Each capability row shows the capability name and its description.
  • Docs — Renders the agent's extended documentation as formatted markdown.

Installing an agent​

Beak Hub is the catalogue. Installation happens from the Runtime that will host the agent, not from here.

To install an agent, open Marketplace in the Beak AI Runtime for the target tenant, choose the Agents tab, and install it there. The agent is created in that Runtime and from then on behaves like any other agent in it — it appears in the Runtime's agent list, is configured there, and is governed by that Runtime's approval policies.

Note: Earlier versions of Hub carried a Deploy to Runtime action on this page, with the destination Runtime configured in a Hub settings screen. That flow has been replaced: the Runtime now pulls from the Hub catalogue rather than Hub pushing into a Runtime. There is no deploy action, and no Runtime URL to configure, in Hub today.

Invocation Modes​

Agents declare the channels through which they can be invoked. The badges on the catalog card and detail header reflect these modes:

  • Chat — The agent is accessible via the Runtime chat interface.
  • API — The agent exposes a direct API endpoint.
  • A2A — The agent supports the Agent-to-Agent protocol for machine-to-machine calls.
  • Copilot — The agent can be published as a Microsoft Copilot agent via passthrough mode in the Runtime.
  • Teams — The agent can be exposed as a Microsoft Teams bot.
  • Studio — The agent can be imported into Beak AI Studio.

Best Practices​

  • Review the Capabilities tab before installing an agent to confirm it covers the use case you need; capability descriptions explain what each function does.
  • Install into a test Runtime before rolling out to production tenants; an installed agent is immediately available to everyone who can see it.
  • Agents that support A2A can be chained with other agents in Beak AI Studio workflows without requiring a human-in-the-loop.
  • The Docs tab often contains prerequisite configuration steps for skills, datasources, or MCP servers the agent relies on — read it before installing, to avoid credential gaps in the Runtime.
  • Agents are installed from the Runtime's Marketplace, not from Hub. If the Marketplace is not visible there, the operator lacks the marketplace install permission.
  • Overview — Beak Hub landing page and catalog summary.
  • MCP Servers — Pre-built MCP servers that agents consume as tools.
  • Agent Suites — Curated bundles that include multiple agents alongside skills and datasources.
  • Marketplace — Where agents from this catalogue are installed into a Runtime.