nervous-system

extensionmaintained

NERVous System — durable multi-agent orchestration, coding-agent coordination, and workflow recovery for pi.

by · v1.9.2 · published 1w ago

$ pi install npm:nervous-system
downloads/mo
4.2K
stars
1
last push
1w ago
open issues
1

Signals

license: MITtestspi manifest: missinginstall size: —deps: 0peer deps: 0

Download trend

4.3K downloads · last 12 weeks (weekly)

README

NERVous System

A durable multi-agent orchestration and coding-agent coordination framework for pi. NERVous System helps coding agents plan, delegate, coordinate, execute, recover from interruptions, and review work continuously.

Invoke /nervous, give it a task description, and walk away.

Find it in the pi package directory by searching for multi-agent orchestration, coding-agent coordination, subagent delegation, durable task planning, workflow recovery, or risk triage.

Quick start

pi install npm:nervous-system

Then start a durable workflow:

/nervous implement this feature with durable planning and worker delegation

NERVous installs the complete suite: CORTEX, MAGI, AXON, SYNAPSE, LION, CEREBEL, GANGLION, AMYGDALA, and the dashboard. Its model tools are inactive in a fresh session chain. An exact /nervous [task] command activates the operator-permitted component set after any current run becomes idle, then keeps the coordinated workflow available for later turns on that branch. Keep the package installed but turn its runtime resources off or back on with /nervous:config enabled=false or /nervous:config enabled=true. Use /nervous:state to see exactly what the current work context retains and /nervous:reset to archive that whole context before unrelated work.

Highlights

  • One-time task framing turns abstract requests into concrete goals before deliberation
  • Durable goals and task state that survive compaction, restarts, and interruptions
  • Pi-native transport retries with an explicit settled-interruption resume valve
  • Deliberation, risk triage, and explicit verification for difficult work
  • Parallel coding subagents with orchestration, capacity allocation, and live progress
  • Project- and context-isolated state with explicit retention diagnostics, safe reset archives, and a fault-isolated dashboard

Documentation

GuideContents
Documentation indexSuggested reading paths and all guides
Getting startedInstallation, first use, updates, removal, and local development
ArchitectureComponents, execution flow, repository layout, and package design
OperationsWorker controls, dashboard, state isolation, and configuration
BenchmarkNERVous prompts compared with a raw LLM baseline
Release policyVersion compatibility, state policy, and known limitations

Status: The complete core component set and deterministic end-to-end demo are implemented and tested.