Skip to main content

Overview

TrueFoundry Agent Hub is a centralized platform for building, registering, discovering, and orchestrating AI agents within an organization. This will support building complex agents as well as registering and using pre-existing agents.

Capabilities

Agent Hub provides the following capabilities:
  1. Agent development using models and MCP servers Build agents using foundation models and MCP servers integrated with TrueFoundry, and publish them as first-class agents within the platform.
  2. Agent apps for non-technical users Expose agents as Agent Apps for use by non-developer or business users, enabling organization-wide access to AI-powered workflows.
  3. Multi-agent workflows with sub-agents Create complex agents by composing and orchestrating sub-agents to model sophisticated multi-step or collaborative workflows.
  4. Bring your own A2A agents Integrate externally built A2A compatible agents and use them as sub-agents.
  5. Agent sharing and orchestration Share agents across teams and orchestrate them into reusable workflows to improve collaboration and reduce duplication.

Upcoming Features

The following enhancements are planned for upcoming releases:
  1. Backend context management APIs APIs to support multi-turn conversations by passing previous response IDs, simplifying state management for agent integrations.
  2. Automatic summarization and smart context handling Intelligent summarization of long conversations to manage context length efficiently while preserving conversational quality.
  3. Large tool-call handling Automatic chunking and management of oversized tool-call responses to ensure reliable and efficient LLM processing.
  4. Skills library A reusable skills framework that allows teams to upload and manage shared skills, which agents can invoke on demand.
  5. Task management integration Built-in task management at the gateway level to support long-running and asynchronous agent tasks.

Building and Publishing Agents

This guide describes how to build and publish agents on TrueFoundry using Agent Hub. Every agent on Agent Hub is backed by a TrueFoundry Prompt. To publish an agent:
  1. Create and experiment with a prompt in the TrueFoundry Playground.
  2. Publish the prompt as an agent on Agent Hub.

Building an Agent

To build an agent, start with the TrueFoundry Playground:
  1. Select a model
  2. Add system prompt and user prompt
  3. Select MCP Servers (one or more from the list of MCP Servers available on TrueFoundry)
  4. Run the prompt to get responses
TrueFoundry Playground Interface
TrueFoundry Playground full view
After experimentation, save the prompt. For details on saving prompts, see the Prompt Management section. For MCP Servers, refer to AI Gateway.

Publishing an Agent

Once the prompt is ready, publish it on Agent Hub from either of the following locations:
  1. On the Agent Hub page, click the “New Agent” button and select the prompt to publish.
Creating a New Agent from Agent Hub
New Agent page
  1. On the prompts page, click the “Publish” button.
Publishing an Agent from Prompts Page
Publish button on prompts page
In both cases, provide the following details:
  1. Name: A unique name for the agent (unique within a company). Every agent is identified by its name.
  2. Description: A description for the agent.
  3. Source: Select the prompt from the list of available prompts.
  4. Collaborators: Add collaborators to the agent. “Manager” role can view and make changes to the agent. “Viewer” access can be granted to end-consumers of this agent.
  5. Sample Inputs: Provide sample inputs for the agent. These are displayed to end-consumers on the “Agents Use” page.
Agent Publish Form
Agent publish form

Using via UI

Once published, all registered agents are visible on the “Agents Hub” page. Click the “Try Now” button to interact with an agent. Try Now Button on Agent Hub
Try Now button on agent
This opens a standalone page where users can interact with agents they have access to. Standalone Agent Interaction Page
Agent standalone page

Integrate with Code

All changes are API-driven. The API contract is currently under active development and will be finalized and documented in the coming weeks.

Using Sub-Agents (Beta)

Enable sub-agents in TrueFoundry by selecting the “sub-agents” option in the playground. Sub-Agents Option in Playground
Sub-agents option in playground
Note: Only one agent can be active at a time. When one agent triggers another agent, the first agent is paused and the second agent is activated. Users will see an “Agent Switch” message in the UI indicating which agent is active. Once the sub-agent execution completes, it returns the final response to the parent agent.