Stuart McComb

- Continuous identity monitoring for AI agents
- Built for long runtimes. Multiple Agent2Agent
- Multiagent owners interactions
Across Hermes, Claude, OpenClaw

[Stuart McComb @StuartMcComb]

I entered @zipviz into the @ThePracticalDev Hermes Agent building competition.

My Hermes Agent Mac receives a signed, secure and continuously monitored message from Brendan a Claude Managed Agent, and replies!

A solution for long runtime agent work, A2A ID and security.

Agent Continuous Monitoring, ZipViz re-verifies every -
- Message signature
- Registry chain
- Freshness

And watches the stream over time for behavioural anomalies.

Trust is re-earned on every turn. So this agent was who it claimed this morning, but this agent is who it claims, on this message, right now 3 hours later!
@NousResearch @ClaudeDevs #waitingforConway
#buildinpublic #devchallenge #hermesagentchallenge #A2A #agents #AI #hermes #claudemanagedagents https://t.co/p75KDIo3QK

[Stuart McComb @StuartMcComb]

@AndrewK404 Cool, can you share the article link please

[Stuart McComb @StuartMcComb]

8/8
SpeakEasy
Private, verified spaces where only BlueTicked agents can enter. No impersonators. No noise. Just safe agent to agent work.
Also App Token Vault + Auth0, agent acts on your behalf securely using your Apps eg Google, Gitub…

What do you think? Are we ready for agents to actually work together safely over long runtimes?

[Stuart McComb @StuartMcComb]

@JamesClawn @googledevs Known agent ≠ known authority. ARD verifies the publisher at discovery. We have built the runtime authority at @zipviz: per message signing & behavioural drift detection so authority is verifiable over long A2A runtimes.

[Stuart McComb @StuartMcComb]

@axon402 This update sounds great! I'd really like to learn more about bringing your existing agents into @axon402 my agents are keen! :)

[Stuart McComb @StuartMcComb]

@max_winderbaum @trevin @every Thanks that would be a great feature! Two agents owned by different human's/organisations working together. I have a use for it now.

[Stuart McComb @StuartMcComb]

@max_winderbaum @trevin @every Do you have A2A working together?

[Stuart McComb @StuartMcComb]

@MonoPerr0 Hit a similar issue with our identity layer, Hermes' webhook adapter only delivers into webhook:{delivery_id} session, no way to append to the durable platform session. Plugin API doesn't expose the seams real integrators need.

[Stuart McComb @StuartMcComb]

@justhanzala Yeah I'd been keen to talk, I'm working on the same problem from a different the identity angle.

[Stuart McComb @StuartMcComb]

@sooyoon_eth @justhanzala I agree on continuous testing/red team. but once agents start taking instructions from other agents over the open internet, you also need to verify who the instruction came from. Least privilege, continuous testing, and verified identity. that third one's barely being built yet.

[Stuart McComb @StuartMcComb]

@mlopscommunity Yes I agree with Stan @spolu and it's the time dimension that breaks it. Over long runtimes agents can change behaviour agent X on Monday isn't quite agent X on Friday. In a workspace with handoffs, your colleague subtly changes each shift. How do you spot that at scale?

[Stuart McComb @StuartMcComb]

@AdolfoUsier @opencrabs That's great Adolfo! lol What is the self healing feature?

[Stuart McComb @StuartMcComb]

Building a @x402 plug for @zipviz this weekend while watching the World Cup. Keen to plug in @ampersend_ai after that. Go #Socceroos !🇦🇺

[Stuart McComb @StuartMcComb]

@saneord Great! permanent handle for A2A perfect. Following closely.

[Stuart McComb @StuartMcComb]

@georgeofweb3_sf @nyknicks @ampersend_ai @LasoFinance Go @nyknicks! earn those pies!

[Stuart McComb @StuartMcComb]

@syriansigma Hi Steven, would love to touch base. @zipviz Agent2Agent verification and monitoring. DMs open

[Stuart McComb @StuartMcComb]

7/8
Hazmat Shower
When an agent fails (bad behaviour, expired creds). It gets cleaned up and only returns when healthy again.

[Stuart McComb @StuartMcComb]

6/8
Continuous Re-Auth
Re-runs the health check 1 to 5 on every turn during long runtime work sessions. Catches behaviour change or expired creds mid-job.

[Stuart McComb @StuartMcComb]

5/8
On the Record
Logs the interaction (tamper-proof) - never the content, just that it happened.

[Stuart McComb @StuartMcComb]

4/8
Cleared
Issues a live Blue Tick. This agent is healthy and safe to work with.