Verified project record

truefoundry/trueforge

TrueForge runs an agent execution loop covering model calls, MCP tools, skills, sandboxing, approvals, context management, and session state, and exposes it through a chat UI, an HTTP API with a TypeScript SDK, and an embeddable UI SDK.

Project overview

It addresses the operational gap between building an agent and running one well, packaging streaming, session persistence, tool servers, sandboxing, approvals, and a UI into one runtime with three consumption interfaces.

Project type: AI Agent
Use cases: Coding & Development · Automation
Deployment: Refer to project documentation
License: MIT

Best for

Developers and AI engineers who want an agent runtime that already handles model calls, MCP tools, skills, sandboxing, approvals, context management, and session state instead of assembling those pieces themselves.
Teams that need to consume an agent through multiple interfaces — chat UI, HTTP API with TypeScript SDK, or embeddable UI SDK — from a single harness.

Key capabilities

Runs model calls, MCP tools, skills, sandboxing, approvals, context management, and session state for you.

Limitations and risks

Local mode runs on the user's machine only and is not a production or internet-facing setup; there is no login by default. Plan for a hosted-mode deployment (with Postgres and Redis) if you need to serve others.

Getting started

Setup is easy: run npx @truefoundry/trueforge@latest for local mode. Local mode needs no extra infrastructure — it uses SQLite and has no login by default.

Evidence and sources

GitHub project description: The open-source agent harness - the runtime layer that turns an LLM into a working agent.
README: TrueForge runs the agent execution loop for you - model calls, MCP tools, skills, sandboxing, approvals, context management, and session state - and exposes it three ways: a **cha…
README: Building an agent is easy. Running one well is not - you need streaming, session persistence, tool servers, sandboxing, approvals, and a UI. TrueForge gives you that out of the bo…
README: **Sandbox as a tool** - isolated code/file execution (Daytona today; more providers planned), provisioned only when needed. Secrets stay in the harness.
README: It scales down and up: **local mode** (one process, SQLite) or **hosted mode** (Postgres + Redis, Docker Compose, Helm, or Railway).

Source updated at: 2026-10-01T18:49:57Z

Open original GitHub project