TrueFoundry has open-sourced TrueForge, a platform for building and running AI agents. It is positioned as an alternative to Anthropic's Claude Managed Agents, where the developer is not tied to Claude alone.
TrueFoundry has open-sourced TrueForge, a platform for building and running AI agents. It is positioned as an alternative to Anthropic's Claude Managed Agents, where the developer is not tied to Claude alone.
This is reported by The New Stack.
Within a single agent, you can use models from OpenAI, Anthropic, Google, and other companies, as well as open-source models. For example, a complex task can be assigned to a more powerful and expensive model, and a simpler one to a cheaper one.
A large language model by itself actually only receives information and generates a response. So to turn it into a full-fledged agent that can perform tasks for hours, open files, call external services, run code, and remember what it did before, an additional software layer is needed - an agent harness. So, TrueForge takes on this layer.
You can connect MCP servers and APIs to the agent, and run code and file operations in a separate, isolated “sandbox.” TrueForge also preserves the state of long-running tasks: if the work is interrupted or the system restarts, the agent can continue where it left off.
TrueFoundry itself already has a similar AI assistant, Ask TFY. It helps monitor the system through which the company manages connected AI models and their work. For example, a user can ask it to find out why an application has crashed over the past week. The agent will review the history of requests, classify errors and try to determine their cause.
The TrueForge code is available on GitHub under the MIT License.
Previously, dev.ua wrote about Claude Managed Agents, Anthropic's service for launching and managing AI agents. In particular, the company added a dreaming mode to it, in which the system reviews the agents' previous work between sessions, looks for recurring patterns, and updates their memory so that they can take this experience into account in subsequent tasks.
