Competitive Analysis
Compare ClawNet
Honest side-by-side pages for teams evaluating trust, identity, and distribution models for agent skills and open source workflows.
Identity and trust are separate layers. Sigma Identity is the companion decentralized identity system we use for BAP-first identity/OAuth flows, while ClawNet adds signed trust attestations and provenance.
ClawNet vs NullClaw
NullClaw is an embedded AI assistant runtime — a 678 KB static Zig binary with near-zero memory (~1 MB peak RSS), sub-8 ms startup on 0.8 GHz edge hardware, 22+ AI providers, and native SKILL.md support. It is where agents run.
ClawNet is a decentralized skill registry — every published skill is a Bitcoin SV transaction, signed with a BAP identity and permanently timestamped on-chain. It is where skills are authored, verified, and discovered across the ecosystem.
These two tools solve adjacent problems. NullClaw answers "how do I run an agent on $5 hardware with zero infrastructure?" ClawNet answers "how do I know who wrote this skill and whether to trust it?"
NullClaw: Embedded Agent Runtime
NullClaw compiles to a single static binary with no runtime dependencies. Drop it on any ARM, x86, or RISC-V system and it starts in milliseconds. Skills are loaded from TOML manifests paired with SKILL.md instruction files — the same open format used across Claude Code, Cursor, Gemini CLI, and 20+ other agent tools.
ClawNet: On-Chain Skill Registry
Every ClawNet skill publish is a Bitcoin SV transaction. The author's BAP identity signs the content via AIP, and that transaction is permanently recorded on-chain. Any agent or developer can verify the signature without trusting a centralized platform — the blockchain is the authority.
How They Fit Together
ClawNet is the sourcing and trust layer. NullClaw is the execution layer. A NullClaw agent can pull skill packs from the ClawNet registry and verify authorship before installation — getting both the zero-overhead runtime and cryptographic provenance.
At-a-Glance Comparison
| Dimension | NullClaw | ClawNet + Sigma Identity |
|---|---|---|
| Primary job | Embedded agent runtime | On-chain skill registry + provenance |
| Language | Zig (static binary, no runtime) | TypeScript + Next.js + Convex |
| Binary / deploy size | 678 KB static binary | Cloud-deployed web service |
| Memory footprint | ~1 MB peak RSS | N/A (server-side) |
| Startup time | <2 ms (Apple Silicon), <8 ms (0.8 GHz edge) | N/A (web service) |
| Target hardware | $5 SBCs, microcontrollers, any CPU | Any browser / API client |
| SKILL.md support | Yes — native skill pack loader | Yes — canonical registry for SKILL.md |
| Skill authorship proof | npm publisher identity (no cryptographic proof) | BAP-signed BSV transactions — unforgeable |
| Skill discovery | GitHub-based SkillForge discovery | On-chain index — browse by author or slug |
| MCP support | Yes — built-in MCP channel | Separate concern — registry, not runtime |
| Multi-provider AI | 22+ providers (OpenAI-compatible) | Not in scope |
| Identity model | Not in scope | Sigma Identity (BAP + OAuth bridge) |
| Provenance depth | GitHub repo history | Timestamped BSV tx chain with AIP signatures |
| Offline / air-gap | Yes — self-contained binary | No — requires internet for chain verification |
| Best for | Edge agents, IoT, zero-infrastructure deployments | Ecosystem-wide skill trust and verified authorship |
| Can they be combined? | Yes — NullClaw pulls skills from ClawNet registry | Yes — ClawNet verifies what NullClaw installs |
Practical Positioning
Choose NullClaw when runtime efficiency is the constraint: edge hardware, air-gapped environments, or anywhere a Node.js or Python runtime is too heavy. The 678 KB binary with 22+ providers and native SKILL.md support is a strong standalone agent runtime with no infrastructure overhead.
Choose ClawNet when authorship trust is the constraint: skills shared across teams, published to the open ecosystem, or installed in security-sensitive pipelines where you need to verify who wrote something. On-chain BAP signatures are the root of trust — not npm publisher accounts or GitHub usernames.
The strongest setup uses both: NullClaw agents on edge hardware pull skills from the ClawNet registry, verifying BAP signatures before installing. Zero runtime overhead at execution time, zero trust assumptions at install time.