A governed multi-agent society built upon the principles of MindForth,
the first working symbolic AGI for autonomous robots.
Activation spreading · Associative memory · Self-referential inference · Constitutional governance
"The ghost in the machine is not a single mind — it is the emergent intelligence of a governed society."
Ghost Parliament evolves the original MindForth symbolic AGI — 6,735 lines of Forth representing decades of research into autonomous machine intelligence — into a modern, governed multi-agent system. Every principle of MindForth is preserved and elevated:
| MindForth Principle | Modern Realization |
|---|---|
| Psy Array (concept grid with 21 associative tags per thought) | Structured agent memory with typed slots, retrieval, and decay |
| SpreadAct (activation spreads from active concepts to associated ones) | Event-driven message passing between agents, weighted by relevance |
| InFerence (transitive reasoning via be-verb chains) | Formal proposal lifecycles and deductive governance rules |
Self-referential thinking (the isolation counter triggers Imperative when lonely) |
Agents autonomously initiate proposals, raise concerns, and question |
| ReJuvenate (memory recycling at the edge of the concept grid) | SQLite-backed persistence with archival, pruning, and replay |
| PsiDecay (conceptual activation decays over time) | Attention-scoped message routing with time-to-live |
| MindBoot (primordial knowledge base loaded at startup) | Constitutional articles and foundational governance rules |
| ConJoin (conjoining multiple active thoughts into one utterance) | Multi-agent debate synthesis and consensus formation |
| MainLoop (perceive → think → act → recycle) | Sense → Propose → Deliberate → Vote → Enact → Learn |
Ghost Parliament provides infrastructure for creating, simulating, and governing multi-agent societies. It packages the CAMEL-AI multi-agent framework with a formal governance layer inspired by the architectural patterns of MindForth:
- Symbolic Core — Activation-spreading associative memory with typed concept slots, recall vectors, and time-based decay
- Parliament — Proposal-driven governance with voting, constitutions, reputation, and auditable decision trails
- Simulation — Multi-agent debate, session orchestration, and sandboxed worlds for long-running experiments
- Memory — SQLite persistence with append-only ledgers, replay audit, and cross-session recall
┌──────────────────────────────────────────────────────────────┐
│ USER INTERFACES │
│ CLI · API · Dashboard · Notebooks │
├──────────────────────────────────────────────────────────────┤
│ APPLICATION LAYER │
│ Simulations · Experiments · Research Pipelines │
├──────────────────────────────────────────────────────────────┤
│ SOCIETY LAYER │
│ ┌──────────┐ ┌──────────┐ ┌──────────────────────┐ │
│ │ PROPOSALS│ │ VOTING │ │ CONSTITUTION │ │
│ │ submit │ │ ballot │ │ articles │ │
│ │ review │ │ quorum │ │ amendments │ │
│ │ track │ │ outcome │ │ parser │ │
│ └──────────┘ └──────────┘ └──────────────────────┘ │
│ ┌──────────────────────────────────────────────────────┐ │
│ │ REPUTATION SYSTEM │ │
│ │ ledger · scoring · policy · eligibility │ │
│ └──────────────────────────────────────────────────────┘ │
├──────────────────────────────────────────────────────────────┤
│ AGENT LAYER │
│ Communication · Coordination · Task Execution │
│ (CAMEL-AI multi-agent framework) │
├──────────────────────────────────────────────────────────────┤
│ SYMBOLIC CORE (MindForth Heritage) │
│ Concept arrays · Activation spreading · Association │
│ chains · Inference engine · Query resolution │
├──────────────────────────────────────────────────────────────┤
│ INFRASTRUCTURE │
│ SQLite · Storage · Networking · Security · Monitoring │
└──────────────────────────────────────────────────────────────┘
Architecture → · Integration Guide →
git clone https://github.com/NullLabTests/mindforth-society.git
cd mindforth-society
pip install -e .from parliament import Proposal
proposal = Proposal(
title="Add weather toolkit",
summary="Integrate a weather API toolkit for agent use.",
rationale="Agents need real-time weather data for logistics tasks.",
sponsor="agent_alice",
tags=["enhancement", "toolkit"],
)
proposal.submit()
print(f"Status: {proposal.status.value}")from parliament import Ballot, Vote, VotingSystem
vs = VotingSystem()
vs.cast(Ballot(proposal_id="abc123", voter_id="alice", vote=Vote.YES, weight=2.0))
vs.cast(Ballot(proposal_id="abc123", voter_id="bob", vote=Vote.NO, weight=1.0))
outcome = vs.outcome("abc123", total_electorate=5)
print(f"Passed: {outcome['passed']}")from parliament import Constitution, Article
constitution = Constitution(preamble="We the agents...")
constitution.add_article(Article(
title="Voting Rights",
body="Every agent with positive reputation may vote.",
section="rights",
))from parliament import ReputationLedger, ReputationEvent, ReputationEventType
ledger = ReputationLedger()
ledger.record(ReputationEvent(
agent_id="alice",
event_type=ReputationEventType.PROPOSAL_PASSED,
delta=20.0,
rationale="Passed a proposal",
))
print(f"Alice's reputation: {ledger.score('alice')}")python -m pytest test/parliament/ -v| Suite | Tests | Status |
|---|---|---|
| Parliament (proposals) | 10 | ✅ Passing |
| Parliament (voting) | 8 | ✅ Passing |
| Parliament (constitution) | 5 | ✅ Passing |
| Parliament (reputation) | 8 | ✅ Passing |
| Memory | — | 🚧 In progress |
| Simulation | — | 🚧 In progress |
| Worlds | — | 🚧 In progress |
| Version | Theme | Status |
|---|---|---|
| v0.14 | 🏛 Governance & Auditability | ✅ Current |
| v0.15 | 🧪 Agent Society Simulation | 🔄 Planned |
| v0.16 | 💾 Memory & Persistence | 🔄 Planned |
| v0.17 | 🌐 Multi-Society Federation | 📋 Planned |
| v0.18 | 📊 Analysis & Visualization | 📋 Planned |
| Goal | |
|---|---|
| 🧠 | Emergent Governance — How do governance structures emerge without central coordination? |
| 🎲 | Reputation Dynamics — What mechanisms produce stable, cooperative agent societies? |
| 🛡️ | Constitutional Robustness — How resilient are rules under adversarial behavior? |
| 🤝 | Cross-Society Coordination — How can societies form treaties and resolve conflicts? |
| 🧑⚖️ | Human-in-the-Loop — What interfaces enable effective human oversight? |
| ⏱️ | Long-Running Experiments — What dynamics emerge over weeks or months? |
| 👻 | Symbolic Grounding — Can activation-spreading architectures ground abstract symbols in experience? |
Contributions of all kinds welcome — code, docs, governance proposals, and experiments.
| Area | Area | ||
|---|---|---|---|
| 🐛 | Report bugs | ✨ | Propose features |
| 🏛 | Governance proposals | 🔬 | Research proposals |
| 📖 | Improve docs | 🧪 | Write tests |
| 🤖 | Contribute your agent | 🗳️ | Vote on proposals |
Apache License 2.0 — Fully open, community-governed.
Ghost Parliament — Symbolic AGI · Governed Multi-Agent Societies · Emergent Intelligence
Built upon MindForth (1999–2019), the first working symbolic AGI.
Join the Parliament.