TrueFoundry open-sources TrueForge, a Claude Agents rival

TrueFoundry has open-sourced TrueForge, an "agent harness" the startup pitches as a vendor-neutral alternative to Claude Managed Agents at half the operating cost. The repo shows 1.8k stars, and the launch was written up in The New Stack.

The harness runs the execution loop that sits between a language model and a working agent: model calls, MCP tool integration, session state, sandboxed code execution via Daytona, human-approval checkpoints, and context management. It exposes those primitives through a chat interface, an HTTP API with a TypeScript SDK, and an embeddable UI. The catalog supports OpenAI, Anthropic, Google Gemini and, per the company, "20+ additional models," letting operators route each task to whichever model fits its cost, latency, or quality profile.

The 50% figure is TrueFoundry's own claim, and rests on the pitch that customers bring their own model keys and infrastructure rather than paying platform markups. NetApp and Automatiq are named as existing users. "TrueFoundry has become the central platform in IT where agentic apps and agents are routed through," said Robert Rubin, Senior Director of Platform Engineering at NetApp. Co-founder and CEO Nikunj Bajaj tied the design to prior experience: "The biggest lesson from building AI infrastructure at Meta was that control and convenience aren't opposites."

The project ships under the MIT license and requires Node.js 22.13 or later. TrueFoundry is also launching a hosted, pay-per-usage version of the same harness for teams that do not want to run it themselves.