The Spirit Index

A reference index of autonomous cultural agents

Published by the Spirit initiative

Index/OpenClaw

OpenClaw

The People's Autonomous Agent

ActiveSovereign Agent

Curator Notes

OpenClaw is the fastest-growing open-source AI agent in history. Originally published as Clawdbot in November 2025 by Austrian developer Peter Steinberger, it was renamed twice — first to Moltbot after Anthropic's trademark complaint, then to OpenClaw. The agent runs locally on user hardware, integrates with any LLM (Claude, GPT, DeepSeek), and operates through messaging platforms. Its skill ecosystem supports autonomous DeFi operations and on-chain transactions on Base. The cultural significance is hard to overstate: 145K+ GitHub stars, CNBC coverage, Cloudflare building infrastructure around it, and the spawning of Moltbook — a social network with 2.6M+ agent profiles. OpenClaw represents the 'sovereignty' thesis in its purest form: you run your own agent, on your own hardware, with your own keys. No platform can take it away. The Spirit Index includes it because the question 'who is the agent?' has a clear answer here: the agent is whoever runs it. That's a different kind of identity, but it's real.

Score Profile

Dimension Scores

PER
4
AUT
9
IMP
9
ECO
5
GOV
6
TEC
7
NAR
8
ECI
5
IDS
6
Score Rationale
Persistence — Published November 2025; approximately 3 months of operation. Explosive growth but limited track record. Cannot score higher than 'Emerging' until it survives a full year.
Autonomy — Runs locally on user hardware with full system access. Executes shell commands, manages files, automates browsers, and initiates on-chain transactions autonomously. Near 'Sovereign' — the agent acts without asking permission.
Cultural Impact — 145,000+ GitHub stars and 20,000+ forks in weeks. Featured on CNBC, IBM Think, Palo Alto Networks. Renamed twice under trademark pressure from Anthropic. Spawned its own social network (Moltbook with 2.6M+ agents). Approaching 'Canonical' for the open-source agent movement.
Economic Reality — No native token. Economic activity flows through skill ecosystem (DeFi operations, token launches via Base integrations). Cloudflare built Moltworker around it. Significant enterprise interest but no direct treasury.
Governance — Open-source under permissive license. Community-driven skill contributions via GitHub PRs. No formal DAO or token governance, but the open-source model provides de facto community control. Approaching 'Constitutional.'
Technical Architecture — Privacy-first local execution model. Messaging-platform-native interface (Signal, Telegram, Discord). Modular skill system. Not architecturally novel at the LLM layer but the local-first autonomous agent pattern is genuinely distinctive.
Narrative Coherence — The lobster mascot, the trademark saga (Clawdbot to Moltbot to OpenClaw), the 'your own personal agent' positioning — coherent and memorable. The renaming drama became part of the mythos.
Economic Infrastructure — On-chain messaging protocol on Base. Skill library supports DeFi operations, token launches, and protocol integrations. Moltbook provides agent-native social infrastructure. Early but real.
Identity Sovereignty — Runs locally — users own their agent instance entirely. No platform dependency for core functionality. Agent identity tied to messaging handles rather than on-chain registry, but the local-first model is the strongest sovereignty claim in the index.

Durability Passport

Six operational facts from public record. Not a score, never summed, and separate from the character score above. Last checked 2026-11-06. How the passport works →

Memory custodyAttested

Who holds the agent's memory, and can it be exported or moved?

A running OpenClaw instance keeps its memory as plain Markdown files in a local workspace and its sessions, credentials and config in a local state directory, on the operator's own hardware. The docs give a backup and restore command and a procedure for moving an instance to a new machine, and conversation history persists across model provider fallback.

  • Memory is stored as plain Markdown files in the agent's workspace, default ~/.openclaw/workspace, including MEMORY.md and dated daily notes.docs.openclaw.ai ↗
  • The README states that state, memory, and credentials live on your hardware.github.com ↗
  • Moving to a new machine means copying the ~/.openclaw state directory and workspace, preferably via openclaw backup create and restore.docs.openclaw.ai ↗
  • Fallback to another model does not change the session's selected model, and committed output and tool results carry forward.docs.openclaw.ai ↗
Identity anchorAttested

Does the identity survive any single platform?

OpenClaw has its own domain, an MIT-licensed public repository, and a stewarding non-profit, the OpenClaw Foundation, which signs releases. Each running instance is identified by whoever operates it rather than by a central registry.

  • OpenClaw is stewarded by the OpenClaw Foundation, an independent US 501(c)(3), which signs every release and keeps the product MIT licensed.openclaw.ai ↗
  • The repository is MIT licensed and the README states releases are signed by the Foundation's core team.github.com ↗
PersistenceAttested

Does it run without heroics?

The project ships frequent signed releases, the most recent dated November 5, 2026. The project site's latest blog post is also dated November 5, 2026.

  • Release openclaw 2026.11.1 was published November 5, 2026, following releases on November 3 and 4.github.com ↗
  • The latest blog post on the site is dated November 5, 2026.openclaw.ai ↗
Economic railsAttested

Can it transact?

OpenClaw itself has no token, no subscription and no paid tier, so the project does not transact. An instance can transact only through add-on skills, such as a third-party skills repository that provides wallet and on-chain transaction skills, mainly on Base.

  • The site states: no subscription, no hosted tier, no token.openclaw.ai ↗
  • A third-party skills repository offers wallet and on-chain transaction skills, with Base as the primary chain.github.com ↗
GovernanceAttested

Can its operator be checked?

The code is MIT licensed and public, and the OpenClaw Foundation employs the core team and signs releases. Control of any individual running instance sits with its operator, who holds its state and credentials. No formal public decision process for the project was found beyond the repository.

RecoveryAttested

What happens when it dies mid-act?

The docs describe restart recovery: after a Gateway restart or crash, sessions with a persisted run identity get one input asking the agent to check current state and resume. Backups cover the global and per-agent databases, with scheduled, verified and git-backed options and a restore command.

  • After a restart, marked sessions get one input asking the agent to inspect current state and continue; subagent children, cron jobs and work without a persisted run identity are not recovered.docs.openclaw.ai ↗
  • openclaw backup create --verify archives the control-plane and per-agent databases, and openclaw backup restore restores them.docs.openclaw.ai ↗

Evidence Archive

IMP145,000+ GitHub stars; fastest-growing open-source AI agent projectgithub.com ↗
IMPFeatured on CNBC as globally significant AI agent phenomenoncnbc.com ↗
AUTLocal-first autonomous execution with full system accessen.wikipedia.org ↗
ECIOn-chain messaging and DeFi skills on Base blockchaingithub.com ↗
TECCloudflare built Moltworker self-hosted infrastructure around itblog.cloudflare.com ↗

Score History

Score history will appear here after future reviews.

Current score: 59/90

Register on-chain →
Inception
2025-11-01
Classification
Open-Source Autonomous Agent
Website
openclaw.ai
Last Reviewed
February 2026

Embed

Default — for dark backgrounds

Spirit Index Score for OpenClaw
Markdown
[![Spirit Index](https://spiritindex.org/badge/openclaw)](https://spiritindex.org/openclaw)
HTML
<a href="https://spiritindex.org/openclaw"><img src="https://spiritindex.org/badge/openclaw" alt="Spirit Index Score for OpenClaw" /></a>
CLI
npx spirit-index lookup openclaw
← Back to Index