TrueForge: The Open-Source Agent Harness for LLM-Powered Agents

Summary

TrueForge is an open-source agent harness designed to transform large language models (LLMs) into fully functional agents. It provides the essential runtime layer, handling complex aspects like streaming, session persistence, tool servers, and sandboxing. Developers can leverage TrueForge to build and deploy robust, scalable AI agents with ease.

Use at your own risk

OSRepos shares public repositories for knowledge and discovery only. Any installation, execution, configuration, or use of code from these repositories is the user's own responsibility. Always review the repository, source code, dependencies, licenses, and security implications before running or installing anything. OSRepos is not responsible for issues, damages, or losses resulting from third-party repositories.

Introduction

TrueForge is an open-source agent harness that serves as the runtime layer to convert a Large Language Model (LLM) into a working agent. It manages the entire agent execution loop, including model calls, MCP tools, skills, sandboxing, approvals, context management, and session state. TrueForge provides a comprehensive solution for running AI agents effectively, offering a chat UI, an HTTP API with a TypeScript SDK, and an embeddable UI SDK.

Why Use and Key Benefits

Building an agent is straightforward, but running one efficiently requires robust infrastructure for streaming, session persistence, tool servers, sandboxing, approvals, and a user interface. TrueForge delivers these capabilities out-of-the-box, simplifying agent deployment and management.

Key benefits include:

Initial setup from catalogs: Configure models, MCP servers, skills, and a sandbox once, with agents picking from connected resources. Presets come from customizable YAML catalogs.
Any model provider: Supports OpenAI, Anthropic, Google Gemini, and other catalog providers, or any OpenAI-compatible endpoint.
MCP tools: Remote MCP servers with header authentication or OAuth, including in-chat authorization.
Skills: Git-backed SKILL.md instruction packs, loaded on demand in the sandbox.
Sandbox as a tool: Isolated code/file execution (Daytona today, with more providers planned), provisioned only when needed. Secrets remain within the harness.
Human checkpoints: Features like tool approval, ask-user-questions, and Generative UI in chat.
Context engineering: Includes subagents, deferred tool loading, Code Mode, large-result offloading, and compaction.
Chat UI + SDK: Use the bundled UI, automate with @truefoundry/trueforge-sdk, or embed @truefoundry/trueforge-ui.
Scalability: Scales from local mode (one process, SQLite) for personal use to hosted mode (Postgres + Redis, Docker Compose, Helm, or Railway) for teams and multi-replica deployments.

Installation

Getting started with TrueForge is quick and easy. You can run a local instance using npx:

npx @truefoundry/trueforge@latest

For more detailed installation methods, including Docker Compose, Kubernetes, or Railway, refer to the Quickstart guide on the official documentation. This guide will help you connect models, tools, skills, and build your first reusable agent.

Examples

TrueForge empowers developers to create and manage sophisticated AI agents. The Quickstart guide provides steps to build your first agent, connecting various resources like models, tools, and skills. For advanced use cases, the documentation covers topics such as creating agents with human checkpoints, utilizing harness capabilities like sandboxing and subagents, and integrating with the TypeScript SDK for programmatic control.