🧠 How Blockli Pulse Agents Deliver High-Quality Intelligence at Low Cost

Blockli Assistant pulse agents are designed to help WordPress-based communities, learning platforms, and stores harness AI insights without overspending on token usage. It achieves this through a dual-layer processing model that balances real-time intelligence with periodic summarization, ensuring every decision is context-aware and cost-efficient.

It uses a multi-tiered processing architecture that combines free and low-cost large language models (LLMs) to optimize both cost and performance.

At the core of this strategy are two components:

• OpenRouter’s free models — used for background or low-priority analysis tasks.

• OpenAI’s GPT-5-nano — a low-cost fallback model used when OpenRouter’s API is unavailable.

This hybrid approach ensures that Blockli Shadow agents continues to deliver reliable, context-aware automation while keeping operational costs exceptionally low.

GPT-5-nano Cost Breakdown:

Only key user milestones—such as course completions, purchases, or rank upgrades—are sent to the LLM for reasoning.
All other day-to-day actions are logged in episodic memory for later use.
Each event is processed with the latest summarized user context, ensuring high-quality insights without re-feeding all historical data.

Average token usage per event:

• Prompt: 2,043 tokens




• Completion: 2,113 tokens

• Total cost: ≈ $0.00095 per event

• 1,000 events ≈ $0.95 total

Once a week, Blockli Assistant performs a background summary sweep of all logged events (milestones + non-milestones).
This creates a compact knowledge snapshot of each user’s journey.
That summary is stored in Redis and reused in subsequent milestone events—maintaining depth while minimizing repetitive tokens.

Average token usage per summary:

• Prompt: 1,155 tokens

• Completion: 1,826 tokens

• Total cost: ≈ $0.00079 per summary

• 1,000 summaries ≈ $0.79 total

✅ High-quality intelligence: Every decision is context-aware, using both milestone data and summarized history.
✅ Low operational cost: Average monthly AI spend per 1,000 active users or events is under $1.00 per category.
✅ Scalable efficiency: Memory caching and selective LLM usage ensure consistent performance as your user base grows.

This cost can effectively be $0 if Blockli Assistant exclusively uses OpenRouter’s free models and never needs to fall back to GPT-5-nano.

OpenRouter’s free tier allows 50 requests per day per account. You can raise this limit to 1,000 requests per day by adding a $10 credit to your OpenRouter account.

That credit is never spent when using free models — it simply acts as a good-faith deposit to maintain account reputation. The credit remains intact unless you deliberately invoke a paid LLM model, which Blockli Assistant does not use by default.

Blockli Assistant only falls back to GPT-5-nano, a low-cost model from OpenAI, when the OpenRouter daily quota is exhausted.

Depending on your site’s user activity and event frequency, you may never incur any token costs for Pulse Agent operations.

Pulse Agents form a lightweight subset of the Blockli multi-agent stack. They focus on background, periodic, and classification tasks — such as summarization or analytics — and are powered by free or ultra-low-cost models like GPT-5-nano.

In contrast, Swarm Agents are interactive, real-time agents that engage directly within the chat interface. They leverage high-end models such as GPT-5, OpenAI’s flagship model for coding, reasoning, and complex agentic workflows. Because of their advanced capabilities, Swarm Agents’ activity contributes to your monthly Blockli token usage, while Pulse Agents typically operate at near-zero cost.

In short:

Blockli Assistant uses smart orchestration, not brute force.
By combining milestone triggers, periodic summaries, and in-memory context caching, it delivers powerful insights with minimal or no token cost.