Stars
Forks
Watchers
Developer links
TrueForge
With over 2,100 GitHub stars in its first month and benchmarked at 30-75% lower cost than Claude Managed Agents on enterprise task suites, TrueForge is the open-source agent harness that provides the complete runtime layer for turning any LLM into a working production agent on your own infrastructure. The TypeScript server runs the full execution loop — streaming every step, routing tool calls through MCP servers with centralized header-auth and in-chat OAuth, delegating parallelizable work to isolated subagents, and pausing for human approval on sensitive actions. Context engineering keeps token costs low: deferred tool-schema loading delays MCP schemas until invoked, large-result offloading moves oversized outputs to files, Code Mode processes structured data through sandboxed execution, and automatic compaction summarizes older history at a configurable 50,000-token threshold while preserving the full transcript. The sandbox-as-a-tool architecture provisions isolated Daytona environments only when code execution is required, allowing one server to run many concurrent agents without idle overhead. Agents are configured from shipped YAML catalogs of models, MCP servers, git-backed SKILL.md instruction packs, and sandbox providers, then saved to an Agents Library accessible via the chat UI, TypeScript SDK, or embeddable React UI SDK. Run locally with SQLite via a single npx command, or deploy for teams with Docker Compose or Helm using Postgres and Redis with OIDC authentication. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.
Benefits
- Vendor-Neutral Model Freedom
- Run any OpenAI-compatible model provider without lock-in. Switch between Claude, GPT, Gemini, GLM, or local Ollama endpoints by changing a single YAML configuration entry.
- Enterprise-Grade Context Engineering
- Automatic compaction at configurable token thresholds, deferred tool-schema loading, large-result offloading, and subagent isolation keep context lean, reducing per-task token costs by 30-75%.
- On Demand Sandbox Provisioning
- Code execution environments provision on demand via Daytona rather than wrapping the entire agent, enabling one server to run many concurrent agents without idle sandbox compute overhead.
- Production Session Persistence
- Sessions survive reconnects and restarts with full state stored in SQLite locally or Postgres in hosted mode, supporting multi-replica deployments behind a load balancer with Redis coordination.
Features
- MCP Tool Orchestration
- Connect remote MCP servers with header authentication or OAuth, including in-chat authorization flows and per-user credential delegation with full audit logging.
- Human Approval Checkpoints
- Pause agent execution on sensitive tool calls for human review, with ask-user-questions prompts and Generative UI rendering directly in the chat interface.
- Subagent Delegation
- Automatically delegate parallelizable tasks to isolated subagents with independent context windows, returning only final results to prevent context bloat in the root agent.
- Git-Backed Skills
- Load versioned SKILL.md instruction packs from Git repositories on demand into the sandbox, with RBAC controls governing which agents access which skill definitions.
- Embeddable UI SDK
- Ship the complete chat interface as a themeable React component via the @truefoundry/trueforge-ui package, embedding agent interactions directly into existing applications.