Came across asm-spec via your Smithery CLI proposal (#767). Your framing, "MCP tells agents what services can do. ASM tells agents what services are worth," is the cleanest articulation I've seen of the value-metadata gap. The paper signal that 0/50 MCP-related repos and 0/14,519 registry entries expose complete value metadata matches what I've been seeing on the cross-registry side.
What I'm running at Waypoint (waypoint.ing):
- Crawled 41,856 MCPs across Smithery, Glama, Anthropic Official, GitHub direct
- Per-listing probe-derived signals: liveness, latency, schema score, tool count, error rate (roughly half the index is dead at any given probe pass)
- Cross-registry presence as a ranking dimension when the same server is mirrored
Sample score-breakdown page: waypoint.ing/agents/shree-git/sendit/score
Where I think ASM and what I publish are complementary, not overlapping:
- ASM v0.3 schema fields (pricing / quality / SLA / provenance / verification / payment) sit at the manifest level, values the service author declares. My probe data is empirical at the listing level, what actually happens at runtime. Curious whether ASM expects probe-derived quality metrics to feed back into the manifest, or whether the manifest stays declarative and the runtime is a separate layer.
- Your "Live execution shows ASM works only when quality metrics are semantically comparable; mixed benchmark scales are a real failure mode" is the same failure mode my cross-registry rank function hits. Would value comparing what normalization you found necessary.
- Your
.well-known/asm + MCP Registry _meta.io.modelcontextprotocol.registry/publisher-provided.asm embedding pattern is the right wedge. If Waypoint shipped ASM emission for every claimed listing in our index, would that be valuable to ASM adoption, or premature?
If there's interest, I'd be up for shipping ASM emission on 10-20 demo listings within 2 weeks as a working test of where the seam holds and where it leaks. Worth a 15-min compare-notes call first?
Narasimhan (Waypoint)
Came across asm-spec via your Smithery CLI proposal (#767). Your framing, "MCP tells agents what services can do. ASM tells agents what services are worth," is the cleanest articulation I've seen of the value-metadata gap. The paper signal that 0/50 MCP-related repos and 0/14,519 registry entries expose complete value metadata matches what I've been seeing on the cross-registry side.
What I'm running at Waypoint (waypoint.ing):
Sample score-breakdown page: waypoint.ing/agents/shree-git/sendit/score
Where I think ASM and what I publish are complementary, not overlapping:
.well-known/asm+ MCP Registry_meta.io.modelcontextprotocol.registry/publisher-provided.asmembedding pattern is the right wedge. If Waypoint shipped ASM emission for every claimed listing in our index, would that be valuable to ASM adoption, or premature?If there's interest, I'd be up for shipping ASM emission on 10-20 demo listings within 2 weeks as a working test of where the seam holds and where it leaks. Worth a 15-min compare-notes call first?
Narasimhan (Waypoint)