Proposal Buckets
Category Evolution: This describes planned capability. See the Roadmap for current status.
When you bring any AI, you need to organize proposals from multiple sources. Proposal buckets let you separate, scope, and manage proposals by AI source, user, or purpose.
The Multi-AI Challenge
You can connect any MCP-native AI to your Control Plane. But what happens when you have multiple AIs proposing actions?
The challenge:
- Claude proposes query optimizations
- Codex suggests schema changes
- Your custom agent flags operational patterns
Without organization, your proposal queue becomes noisy. Proposal buckets solve this.
How Buckets Work
Proposal Queue
├── [Claude] Query Proposals
│ ├── Warm query optimization #1
│ └── New filter suggestion #2
├── [Codex] Schema Proposals
│ └── Field type change #3
└── [Custom Agent] Operational Proposals
├── Threshold alert #4
└── Pattern notification #5
Bucket Scoping
Buckets can be scoped at different levels:
| Scope | What It Means | Use Case |
|---|---|---|
| Workspace | All users see the same buckets | Team-wide AI policies |
| User | Individual bucket configuration | Personal workflow preferences |
| Surface | Bucket per execution boundary | Surface-specific AI assignment |
Use Cases
| Scenario | Bucket Strategy |
|---|---|
| Security audit | Separate bucket for security-focused AI, reviewed by security team |
| Code review | Codex proposals in their own bucket for dev team review |
| Operational monitoring | Custom agent bucket for ops team |
| Feature brainstorming | Claude creative proposals, optional review |
| Compliance checking | Regulatory AI proposals, mandatory approval |
Bucket Configuration (Conceptual)
The following shows conceptual configuration. Actual implementation may differ.
buckets:
- name: "Claude Query Assistance"
source: claude
scope: workspace
auto_approve: false
- name: "Codex Schema Suggestions"
source: codex
scope: user
auto_approve: false
- name: "Ops Agent Alerts"
source: custom/ops-agent
scope: surface
auto_approve: false
Governance Remains Unchanged
Buckets organize proposals. They don't bypass governance. Every proposal - from any bucket - still requires your approval.
What buckets DO:
- Organize proposals by source
- Let you prioritize review by bucket
- Enable team-based review workflows
- Support different AI sources in parallel
What buckets DON'T do:
- Auto-approve proposals
- Bypass the governance loop
- Hide proposals from audit trail
Merging Buckets
Sometimes you want proposals from different sources in the same queue:
Merged Bucket: "All Query Proposals"
├── [Claude] Query optimization #1
├── [Codex] Query suggestion #2
└── [Custom] Query pattern #3
The "Bring Any AI" Enabler
Without buckets, multiple AI sources create chaos. With buckets, you get the benefit of multiple perspectives while maintaining clear governance.
| AI Source | What It Contributes | Bucket Strategy |
|---|---|---|
| Claude | Natural language understanding, explanation | General proposals |
| Codex | Code-focused suggestions | Development bucket |
| GPT | Broad knowledge, alternatives | Research bucket |
| Custom Agents | Domain-specific patterns | Ops bucket |
Current Status
| Feature | Status |
|---|---|
| Single proposal queue | SHIPPED |
| Basic proposal review | SHIPPED |
| Source identification | SHIPPED |
| Named buckets | Planned |
| Bucket scoping | Planned |
| Merge/split buckets | Future |
Learn More
- Execution Model - How the governance loop works
- Connect Any AI - Start connecting MCP-native AIs
- Roadmap - What's coming next