[Feature Request] Cost-Aware Execution: Agent automatically selects lowest-cost path with budget controls

[cc3wangdehai-hue]

Problem

Agents currently execute tasks without cost awareness:

1. No cost visibility: Users do not know the credit cost of each operation before execution
2. Suboptimal routing: Agents may choose expensive AI inference when a simple bash command or free API would suffice
3. No real-time cost feedback: Credits are deducted silently, discovered only after the fact
4. No budget controls: No way to set task-level or session-level spending limits
5. No cost estimation: Cannot preview total cost before starting a multi-step task

Proposed Solution

1. Cost Transparency

• Each available action displays estimated credit cost
• Tool descriptions include cost tier (free / low / medium / high)
• Pre-execution cost confirmation for high-cost operations

2. Smart Cost-Aware Routing

Automatic selection of lowest-cost execution path:

Priority: curl/bash (free) → free API (free) → cached result (free)
          → cheap model (low) → standard model (medium) → deep reasoning (high)

• Agent considers cost as a primary decision factor
• Falls back to cheaper alternatives when budget is limited
• User can override with explicit --no-cost-optimize flag

3. Budget Controls

• Session budget: Max credits per conversation
• Task budget: Max credits per individual task
• Daily budget: Hard ceiling per day
• Alert thresholds: Warning at 50%, 80%, 100%

4. Cost Tracking Dashboard

• Real-time credit consumption per operation
• Historical cost trends
• Cost breakdown by tool/model/type

5. Optimization Suggestions

• After task completion: "This task cost X credits. Using Y approach would have saved Z%."
• Periodic optimization reports

Impact

• Users gain control over spending
• Credit efficiency improves significantly
• Trust increases when costs are transparent
• Platform becomes viable for cost-sensitive users and enterprises

Priority

P2 (Medium) - Important for cost-sensitive users and enterprise adoption