NFLTR

NFLTR provides a control plane for orchestrating AI workers in parallel, enabling coordination of agent-authored work across both local and distributed environments. The platform is designed to manage complex workflows where multiple AI agents or workers—such as planners, implementers, verifiers, reducers, and collectors—collaborate to achieve a goal that is decomposed into child tasks. NFLTR supports a range of worker types, including those backed by Claude Code, Copilot, custom MCP servers, and other LLM-backed agents, allowing users to register workers by their role, model family, repository access, platform, capacity, and locality.

The orchestration process begins by attaching a planner that publishes task digests and polls a command queue. Workers are registered based on their capabilities, and the planner delegates tasks, maintaining lineage so that the dashboard can display the state of root, child, and verifier tasks. Operators can interact with running workflows by answering worker questions, approving or rejecting tasks, or aborting them as needed. The platform records explicit outcomes from verifier, reducer, and loop-controller agents, while storing state information such as live digests, event histories, artifact manifests, and recovery metadata to ensure that work can be inspected, resumed, or recovered after interruptions.

NFLTR is suitable for scenarios where a single agent session is insufficient, such as large code changes, multi-actor local runs, human-gated workflows, distributed machine orchestration, long-running operations, and audit trails. Use cases highlighted include splitting work while maintaining objective lineage, running builders and reviewers on a single machine to validate coordination, routing tasks to machines with specific access or hardware, and enabling cross-platform verification by dispatching changes to different operating systems. The platform allows for both local and distributed orchestration, with the choice depending on factors like coordination quality, resource locality, and hardware requirements.

Users interact with NFLTR via a command-line interface to attach planners and workers, while a hosted dashboard provides visibility into task graphs, approvals, agent activity, and workflow status.