[Discussion] Open-source AgentKit-first observability + policy layer (Argus) — feedback welcome

Hi AgentKit team (and community),

Building Argus - an open-source observability + policy layer for on-chain AI agents. AgentKit is the first adapter on the roadmap: normalised event schema, declarative policies (spend caps, allowlists, rate limits, anomaly detection), cross-provider kill-switch. MIT-licensed, local-first, no SaaS, no token.

Applying to the EF ESP Wishlist under "Existing Tooling for Ethereum Developers" with AgentKit as the first supported SDK.

Rather than asking for a call, three async questions the team (or community) can answer in the thread:

1. From your vantage point inside AgentKit, is a neutral runtime observability + policy layer a gap worth filling externally, or is that surface something AgentKit plans to own internally?
2. What policy primitives are you already seeing operator demand for? (So the v0 policy set matches real pain, not whiteboard guesses.)
3. Is a first-party adapter welcome, or would the team prefer Argus stay adjacent?

Either a short "yes, here's the event-model we'd recommend hooking into" or a blunt "no, because X" saves the project time. Not looking for endorsement - looking for alignment.

If yes, happy to share a v0 preview when the adapter is ready (~4 weeks).

Repo: https://github.com/argus-protocol/argus

Kristiyan Petrov (GitHub: @zontak)
argus.watch@proton.me
https://www.linkedin.com/in/kristiyan-petrov-zontak/