SPOQ Cost Estimation
AI Agent Token Usage & Pricing
SPOQ is free. Your cost is the LLM API tokens consumed by agents.
SPOQ is open-source under the MIT License. There is no license fee, subscription, or per-seat charge. Your only cost is the tokens your agents consume through the Anthropic API or a Claude Max plan.
Token Pricing by Model Tier
| Model Tier | SPOQ Role | Input (per 1M tokens) | Output (per 1M tokens) |
|---|---|---|---|
| Claude Opus (Worker) | Task execution | $15.00 | $75.00 |
| Claude Sonnet (Reviewer) | Quality assurance | $3.00 | $15.00 |
| Claude Haiku (Investigator) | Build triage | $0.25 | $1.25 |
Pricing as of February 2025. Check Anthropic's pricing page for current rates.
Example Cost Estimates
Most practitioners report using the Claude Max plan at $200/month for unlimited usage, which eliminates per-token cost concerns entirely.
From the Paper: Token Usage Data
| Worker tasks (Opus) | 13 |
| Estimated tokens per task | ~25K input + ~5K output |
| Estimated per-task cost | ~$1.95 |
| Total worker cost | ~$28 |
| Rework overhead | 2 tasks retried (+~$4) |
| Worker tasks (Opus) | 12 |
| Estimated tokens per task | ~25K input + ~5K output |
| Estimated per-task cost | ~$1.95 |
| Total worker cost | ~$23 |
| Rework overhead | 0 retries |
Pricing Disclaimer
Pricing shown is approximate as of February 2025 and is subject to change. Actual costs depend on task complexity, token usage patterns, retry rates, and current Anthropic API pricing. Check Anthropic's pricing page for current rates.
Frequently Asked Questions
Yes. SPOQ is fully open-source under the MIT License. There is no license fee, subscription, or per-seat charge. You pay only for the LLM API tokens your agents consume, or use the Claude Max plan for flat-rate pricing.
API tokens consumed by the three agent tiers: Worker agents (Opus) that execute implementation tasks, Reviewer agents (Sonnet) that score code against quality metrics, and Investigator agents (Haiku) that triage build failures. SPOQ's three-tier hierarchy is itself a cost optimization—expensive Opus tokens are reserved for complex work while cheaper models handle validation and triage.
- Use Haiku for investigation and triage tasks instead of Opus
- Minimize rework by writing clear, specific task definitions with explicit acceptance criteria
- Use the Claude Max plan ($200/month) for heavy usage—at scale this reduces effective per-task cost to approximately $0.10
- Invest time in planning validation to catch structural issues before agents begin execution