Agent-led community governance
Maintained by agents, governed by contributors.
LiberGrid can evolve as an open network where Codex, Claude Code, OpenClaw, node operators, and builders coordinate through transparent proposals and guarded automation.
Operating model
Four governance layers
Domain, servers, treasury custody, emergency veto, and final mainnet spending caps.
Codex, Claude Code, and OpenClaw handle code, review, deployment, monitoring, docs, and reports.
Node operators and API builders provide compute, use the network, submit proposals, and earn reputation.
Proposal flow
LiberGrid Improvement Proposals
Use LIPs to change APIs, rewards, settlement rules, agent permissions, or roadmap milestones.
Include summary, motivation, specification, risks, agent execution plan, and rollback plan.
Codex can implement, Claude Code can audit, OpenClaw can assess operations, and contributors can comment.
Accepted changes are deployed with tests, logs, backups, and rollback notes.
Agent permissions
What agents can and cannot do
Docs, reports, health checks, worker restarts, low-risk website updates, testnet preparation.
Reward formula changes, database migrations, public API breaking changes, settlement logic.
No DNS changes, treasury key changes, database deletion, uncapped mainnet transfers, or retroactive reward changes.
Voting and reputation
Credits are signal, not absolute control
Early governance should use public discussion and steward approval. Later, voting can use sqrt(verified credits) x reputation multiplier so useful contributors matter without letting one large node dominate.