An open source rival to Claude Managed Agents just launched

TrueFoundry has launched an open source agent harness called TrueForge, positioning itself as an alternative to Claude Managed Agents from Anthropic. The technology enables software engineers to build, deploy, debug, and govern production AI agents on any model or MCP server, potentially reducing operating costs by up to 50%. TrueForge's neutral approach to model selection challenges the prevailing narrative of managed agent platform lock-in, as providers often lock enterprises into a single vendor's models, infrastructure, and pricing.

The harness serves as the strategic control point between the user, the language model, and the systems the agent interacts with. It handles security considerations, routing sensitive tasks through an isolated sandbox and applying budget enforcement, rate limits, and guardrails through TrueFoundry's AI Gateway. By providing enterprises with ownership of these agent layers, TrueForge aims to address the issue of fragmented workflows and lack of accountability that can arise when key decisions are made collectively by unnamed "foo" and "bar" variables.

TrueForge can be run on-premises, allowing organizations to bring their own models, MCP servers, and API keys. This approach enables them to define the agent development life cycle, enforce consistent operating principles across their AI stack, and ultimately create a more uniform and secure managed agent experience.