🔥 Introducing TORCH — Task Orchestration via Relay-Coordinated Handoff



TORCH is a portable, Nostr-based task-locking toolkit for multi-agent coordination — think of it like a KANBAN board for software agents that evolve with your codebase. It lets schedulers and agents coordinate work by publishing/claiming task locks over Nostr relays so tasks can be reliably handed off without central servers.

What TORCH stands for
Task

Orchestration via

Relay-Coordinated

Handoff — a decentralized lock/claim/expire model for agents and schedulers.

How it works (high level)
Scheduler publishes tasks as lockable Nostr events in a project namespace. Agents see the available tasks and attempt to claim them by writing a lock event to the configured relays.

Locks are time-limited (default TTL ~7200s / 2h) and use standardized tags for expiration (NIP-40), so claims auto-expire and races are resolved by relay-coordinated replaceable events. Kind and expiration behaviour are encoded in the protocol (Kind 30078, expiration via NIP-40).

Rosters & namespaces let you scope daily/weekly task groups to your repo; configuration lives in torch-config.json or env vars so TORCH follows your repo’s cadence.

Agents run locally or in the cloud (AGENT_PLATFORM can be set like codex) and use the CLI to lock, check, and list active locks. There’s also an optional static dashboard.

Why you’d use it
Decentralized: no single server dependency — task state travels with the codebase and relays.

Repo-native: torch-config.json and prompts let TORCH evolve with your project’s workflow.

Ideal for multi-agent automation, documentation agents, CI helpers, and anywhere you want reliable, contestable task claiming.

Try it: demo & docs → https://torch.thepr0m3th3an.net/
Source & quickstart → https://github.com/PR0M3TH3AN/TORCH (see README for the lock/check/list examples).