PROJECTSOPEN SOURCE / 02 · CASE STUDY

    TorqueSight

    Process intelligence above the agent framework.

    TorqueSight maps agent systems as business processes, reads run logs to explain health and cost, ranks improvement opportunities and reviews agent safety. It is a local process-intelligence layer—not another agent framework.

    Open application
    APPLICATION / PUBLIC EXPERIENCE
    TorqueSight public application screenshot
    01 / PRODUCT

    A process map of the agent system

    BPMN provides a visual representation of agents and their relationships. Teams can describe a system or inspect an existing one, then reason about the process rather than only reading disconnected traces.

    02 / ENGINEERING

    One engine, three entry points

    The public product offers a skill, a CLI and an MCP server. The same analysis engine is accessible through guided agent interaction, terminal automation or machine-facing tools.

    03 / ENGINEERING

    Observation becomes an improvement loop

    Run-log analysis connects agent health and cost to ranked opportunities for change. The offline viewer brings the process map, operating evidence and review into one place.

    04 / ENGINEERING

    Integration without owning execution

    TorqueSight sits above whatever runs the agents. The public site names Claude, Codex and other agent environments, alongside workflow and telemetry ecosystems. Local execution is central to the product’s positioning.

    DELIVERY / AI-SDLC

    How the work comes together.

    Built through AI-SDLC: AI-assisted implementation paired with human architecture, review and iterative delivery. Across the studio, engagements range from days to a week or a month; no unverified duration is assigned to this project.

    • BPMN process maps
    • Agent health and cost
    • Skill and CLI
    • MCP server