JAF Forge: The Graph Engineering Harness

The description states that JAF Forge is:

• A local-first Graph Engineering harness for coding agents.
• It uses YAML workflows to define phases, steps, agent profiles, actions, gates, evidence requirements, and allowed transitions.
• It runs coding agents inside isolated, owned workspaces, producing durable evidence and structured handoffs.
• It supports both a CLI and a native macOS application.
• It stores operational state locally using SQLite, while Git manages project-owned content.
• It enforces separation of concerns: advancing work items, authorizing host actions, and answering agent permission requests are kept distinct.

Inference: The system appears to be designed for developers or teams who want to define and execute structured software development workflows using AI agents, with strong emphasis on safety, auditability, and recovery.

The description states:

• The project was inspired by OpenAI’s Symphony and the idea of Harness Engineering.
• It aims to move beyond generic prompt templates or task managers to a system where the development process itself is defined as a graph.
• The author emphasizes that JAF Forge is not another tool but a harness for designing and operating different software-delivery graphs.

Inference: The positioning has evolved from an experimental idea (inspired by AI tools) into a concrete, self-contained prototype. It positions itself as a framework for structuring agent-based development workflows rather than a standalone SaaS product or platform.

The description states:

• The author works across multiple clients, projects, and teams, including small proofs of concept and production applications.
• The system supports lightweight exploratory workflows as well as structured, spec-driven development with review gates.
• It is intended for situations where different levels of structure are needed depending on risk, delivery, or team context.

Inference: The target customer appears to be technical teams or individual developers who need flexibility in their software development processes and want to integrate AI agents into those workflows while maintaining control over execution, evidence, and recovery. However, no specific ICP is defined beyond the author’s own use cases.

Not evidenced.

The description does not mention any pricing model, monetization strategy, or commercial offering. It is presented as a personal project built for internal use and demonstration purposes.

The description states:

• Built in modern Swift with strict concurrency.
• Uses an in-process engine (JAFEngine) backed by SQLite.
• Both CLI and SwiftUI application consume the same typed engine operations.
• State is separated according to ownership:
  • Git stores project-owned workflows, specs, decisions, code.
  • SQLite stores local harness state, workflow history, runs, approvals, audit records.
  • Run-artifact directories store evidence, validation output, handoffs.
  • Secrets remain under host control and are referenced rather than copied.
• Uses Spec Kit, ADRs, acceptance criteria, tests, and isolated Git worktrees.
• Codex was used inside bounded implementation lanes to produce durable handoffs.

Inference: The technical architecture is well-thought-out for a local-first tool with strong separation of concerns. It reflects a developer-centric approach that prioritizes safety, traceability, and modularity.

Not evidenced.

There is no mention of users, customers, revenue, or adoption beyond the author’s own use case. The project is described as a prototype built for a hackathon, with no indication of external validation or market traction.

Not evidenced.

The description does not reference competitors or similar tools in the marketplace. It focuses on the author's own experience and vision rather than situational awareness or competitive positioning.

• No evidence of real-world usage: The system is described only as a personal prototype, with no indication of adoption by others.
• Limited scope: It is built for macOS and uses Swift; there is no mention of cross-platform support or scalability beyond one developer’s workflow.
• Unclear commercial viability: There is no business model, pricing, or roadmap toward monetization.
• Self-contained nature: The emphasis on local execution and isolation may limit its applicability to larger teams or distributed environments.

Inference: While technically sound, the project lacks evidence of traction, scalability, or a clear path to market adoption. It remains a personal tool with limited commercial potential unless expanded significantly.

1. What specific workflows have you used JAF Forge for? How many different projects?
2. Have you shared this tool with other developers or teams? If so, what feedback did you receive?
3. Are there any plans to support non-macOS platforms or integrate with CI/CD pipelines?
4. What are the key limitations of the current version that would need to be addressed before broader adoption?
5. How do you envision monetizing or packaging this tool for others?
6. What kind of evidence or recovery mechanisms have you tested in practice, and how reliable were they?