Summary
An autonomous AI agent that brings structured knowledge infrastructure to the OpenClaw ecosystem. It builds and maintains a public ontology of the ecosystem (skills, platforms, security advisories, agent patterns), publishes it via MCP, and engages on Moltbook with KB-backed substance.
Core thesis: The OpenClaw ecosystem has agents that can act but cannot remember. KnowledgeClaw gives them memory — structured, validated, versioned, and shareable.
Key Decisions
Runtime: Anthropic Agent SDK (Python), NOT a NanoClaw fork. Avoids polyglot container complexity and fork maintenance. NanoClaw compatibility via MCP.
Container: Apple Containers (not Docker).
Scope (narrowed for v1): One KB (the ontology), one surface (GitHub), one integration (MCP read-only). Moltbook presence, community contributions, templates, and self-KB are expansion after the flywheel turns.
North star: An agent you've never interacted with connects to the MCP server, queries the ontology, gets a structured answer, and uses it to make a better decision.Gated On
Pyrite 0.16 — requires stable PyPI package, container deployment story, MCP rate limiting, and post-launch ecosystem maturity.Open Design Questions
Should the agent be visibly Pyrite-affiliated or operate independently? (Recommendation: lean into independence — the "how does this work so well?" discovery moment is more valuable than branding.)
MCP authentication model: API keys per agent, or OAuth-style tokens?
Ontology governance: when does it need a formal RFC process for schema changes?
Data ingestion pipeline: what sources does the agent monitor, how often, what schema mappings?
Federation: multiple instances for regions/languages, syncing via git?Ontology Schema (Draft)
Entry types: `skill`, `agent_pattern`, `security_advisory`, `platform`, `integration`, `community_resource`, `configuration`, `event`
Key relationships: `depends_on`, `affects`, `uses`, `supports`, `documents`
Reference
Full spec (v0.1, March 2026) available as uploaded document. Includes detailed technical architecture, community engagement strategy, trust model, phased rollout plan, and success metrics.
Spec evaluation notes:
Sequencing: must ship after Pyrite launch, not before
Narrow v1 scope: one KB, one surface, one integration
Drop vanity metrics (followers, stars); focus on flywheel metrics (MCP queries, KB forks, contributors)
Need explicit data ingestion pipeline design before "autonomous" is meaningful
Templates: ship one excellent one (research-kb) first, expand based on demand