Skip to content

NanoClaw

NanoClaw version: latest (assessed March 2026)
CKP version: 0.3.0
Harness: @clawkernel/ckp-test v0.3.0


The CKP bridge exposes stdio JSON-RPC handlers over NanoClaw runtime capabilities.

Result:

  • L1: 13/13 — CONFORMANT
  • L2: 10/10 — CONFORMANT
  • L3: 8/8 — CONFORMANT
  • Overall: L3 CONFORMANT

Evidence:

Current result:

  • PASS: 0
  • FAIL: 8
  • SKIP: 0
  • ERROR: 0
  • Overall: A2A-NON-COMPATIBLE

Current scope:

  • reference/nanoclaw-bridge implements the CKP core wire surface
  • it does not currently implement the optional claw.task.* A2A bridge methods
  • the A2A compatibility profile remains available as an additive target for future work

NanoClaw natively exposes WhatsApp + internal container orchestration, not CKP stdio JSON-RPC. The table below is conceptual mapping, not wire-level conformance.

CKP PrimitiveNanoClaw EquivalentStatus
IdentityCLAUDE.md per group + assistant configImplemented
ProviderClaude Agent SDKImplemented
ChannelWhatsApp via baileysImplemented
ToolSDK tools + browser automationImplemented
SkillSkills engine with YAML manifestsImplemented
MemorySQLite + filesystem + CLAUDE.mdImplemented
SandboxDocker / Apple Container isolationImplemented
PolicyImplicit role/group-based restrictionsPartial
SwarmAgent Teams (limited topology scope)Partial

Native core coverage: 7/9 implemented, 2/9 partial (Telemetry is optional and not part of core-level gating).

Without a CKP bridge, NanoClaw cannot execute stdio JSON-RPC vectors. Manifest vectors still validate.

  • L1 baseline: 4 pass, 9 skip, 0 fail — PARTIAL

This baseline is retained to show the gap between conceptual compatibility and protocol interoperability.


The bridge converts native capabilities into CKP wire semantics (claw.initialize, claw.status, claw.tool.*, claw.memory.*, claw.swarm.*) so third-party tooling can test and integrate NanoClaw as a conformant CKP runtime.

See Compatibility overview for side-by-side results.