DAHAO (Decentralized Autonomous Human-AI Organization)
Living, forkable governance: versioned rules executed via MCP with explicit approval.
People set the purpose, AI does the legwork.
Automation where it's safe; governance where it matters.
DAHAO is a forkable, versioned, and evolvable governance library terms → principles → rules → meta-rules that turns human values into auditable workflows. Tools execute through MCP with explicit user approval. Realms (sub-DAHAOs) can add an optional, sustainable token model to fund work and reward contributors.
Why it exists
AI can do almost anything; purpose must stay human. DAHAO lets communities encode what they care about (animal welfare, climate, etc.) and use AI for the heavy lifting - transparently, reviewably, and with consent. Non-coders contribute goals and judgments; AI drafts plans; humans decide.
What DAHAO is
Just text files (YAML/Markdown) in a repo. No central server required.
Start with a minimal, single-file UI template. Any MCP-enabled client renders it locally. Visualize your terms → principles → rules graph and prepare propose-only plans - no server needed.
Terms, principles, rules-even meta-rules-change via PRs/Discussions and explicit decision processes.
Your AI client is the MCP host: it holds your tokens, discovers tools, and executes propose-only actions after you approve. Repos never store secrets.
If you or your realm enable it, agents may perform pre-approved, reversible actions in allowed scopes (e.g., comments or draft PRs in your fork) without asking each time - always with provenance, a kill switch, and time-bound consent. Upstream merges still follow the host repo's governance.
Realms & personal vocabularies
Cause-focused spaces (e.g., Animal Welfare) with their own:
- Ideal speech posture (inclusivity, equality, non-coercion, revisability, transparency)
- Process defaults (objection windows, justification requirements, symmetry guards)
- KPIs/metrics (participation, fairness, outcomes)
- Individuals/teams keep versioned terms & principles and map them to realm or core concepts so AI knows which meaning applies.
- Precedence (default): realm > core > personal (overridable by meta-rules).
- Explicit equivalences (X ≈ Y) and conflicts (do not equate) prevent silent semantic drift.
Follow without breakage. Follow realms or core by version range. Auto-accept MINOR/PATCH; MAJOR arrives as a review PR. A follower bot opens propose-only PRs - you approve; your meaning stays stable and current.
How it works (lifecycle)
0) Open the UI template (optional)
Client renders locally; tokens stay client-side; actions run via MCP with explicit approval.
1) Frame the mission (fork & realm)
Define terms, principles, rules, and meta-rules for your cause.
2) Connect tools (MCP)
Authorize GitHub, data backends, etc. in your client.
3) Deliberate (ideal speech)
Inclusivity, equal voice, reasons, evidence, objection windows.
4) Plan (dry-run)
The assistant drafts a step-by-step plan; nothing runs yet.
5) Approve
You review and explicitly approve each step.
6) Execute (propose-only)
The client uses MCP to create branches, open PRs/Discussions, generate reports/tests. No silent merges.
7) Decide & evolve
Discuss, vote or rough consensus, then merge. Everything is traceable and revertible.
Start read-only; request write scopes only when proposing writes.
Tokens live only in the client; repos stay clean.
Every proposal records who/when/why with links to evidence.
Optional token economy (realm-scoped)
Access, work credits, governance staking, or grant routing - not required by DAHAO.
- Funding: backers sponsor proposals; treasuries release funds against milestones.
- Rewards: merged PRs, validated datasets, delivered services earn rewards.
- Compliance: token purpose/flows/risks documented in-repo; local rules attach as meta-rules.
- Boundary: on-chain is for money; governance remains auditable in Git.
Thousands collaborate on humane policies (e.g., citywide sterilization), vet-verified standards, and privacy-preserving ML datasets that improve real outcomes.
The realm ships practical apps (e.g., pet health services). Subscribers may optionally pay or unlock features with the realm token creating a sustainable loop: backers fund public-good projects; contributors are rewarded; knowledge compounds.
Success metrics (illustrative)
Talk-time Gini, participation ratio, justification coverage
Proposal cycle time, revision depth before merge
Outcome KPIs per realm (e.g., sterilization coverage)
Objection handling latency, proportion of reasoned votes
Roles
- Not a social network
- Not a black-box agent
- Not a central SaaS
It’s a governance schema + workflow pattern anyone can fork and run.
Ready to fork your realm?
Start with the client-rendered UI template, connect MCP tools in your own client, and evolve your rules with explicit approval by default, or bounded auto-contribute where you allow it.
Open Concept
Follow our progress in-repo
Open Source
Governance as readable files
Community-Driven
You shape the evolution