DEV Community

Ashutosh Singh
Ashutosh Singh

Posted on

Why Anthropic’s "Agent Teams" is missing the most important part: Governance.

We’ve been shipping multi-agent orchestration since January. Here is why orchestration is the easy part—and why "Trust" is the real boss battle.

Anthropic just announced Agent Teams. It’s a great validation of the category, but while the giants are focusing on the how of connecting agents, we’ve been obsessing over the should.

​I’ve been building Rigovo Teams for months. 37K lines of code, 8 specialized agents, and one Master Agent later, I’ve realized one thing: The hard problem isn’t orchestration. It’s trust.
​1. Memory without Governance is just "Accumulated Hallucination"

​Most multi-agent systems use a shared memory bank. If Agent A hallucinates an API call and Agent B reads that memory, the mistake is now "institutional knowledge."

​In Rigovo, we built flat semantic memory using pgvector, but with a twist: **
**The Quality Gate
. Every memory entry must pass through 24+ deterministic gates—checking for phantom APIs, secret leakage, and structural violations—before it’s allowed to influence the team.
​2. The Separation of Concerns (Coder vs. Reviewer)

​If you give an agent the same context for coding and reviewing, it will inherit its own biases. In our architecture:
​The Coder and The Reviewer have completely separate contexts.
​The Reviewer doesn't care about the Coder's "opinion."
​It uses Rigour (our open-source governance layer) to perform structural analysis, not just LLM-based vibes.
​3. The Audit Trail: Moving beyond the Black Box
​"Why did the AI make this change?" is the most terrifying question in production.

If you can't trace a decision back to the exact agent, the exact timestamp, and the exact quality gate result, you don't have a team—you have a compliance nightmare.

​Why we Open Sourced it
​We believe the future of AI-generated code depends on transparency
.

That’s why Rigour is MIT licensed. It’s already been forked by teams like Alibaba’s iFlow, proving that production-grade AI needs a governance layer.

​Check out the repo here: github.com/rigovo/rigovo-virtual-team

Governance Layer: rigour.run

​I’m excited to see the big players enter the space, but we’re building this for the engineers who need to ship code that actually works in production, not just in a demo video.

Top comments (0)