If you don't see your question here, reach out.
A real-time monitoring tool that runs alongside your AI agent. It tracks per-turn cost, context accumulation, and tells you the exact moment when spawning a fresh session becomes cheaper than continuing. Single HTML file, runs locally, no account required.
The AEL framework applies to any provider with context caching. Currently supported in AEL products: Anthropic Claude, OpenAI GPT, and Grok (xAI). Gemini support is planned. OpenAI delivers the highest savings due to zero cache write cost. Grok 4 matches the standard 10:1 cache ratio — Grok 4.1 Fast uses a 4:1 ratio but AEL still applies.
Not for the Sidecar. It monitors externally and recommends spawn timing without touching your agent. The Dashboard reads session metadata from your existing log files — no changes to agent logic required.
SpawnPoint ChatReset is designed specifically for browser-based AI interfaces. The Dashboard works with API-connected agents (OpenClaw, LangChain, CrewAI, raw API, etc.).
OpenClaw, LangChain, CrewAI, OpenAI Agents SDK, and raw Anthropic, OpenAI, and Grok API adapters are all supported. If your framework writes session data to a log file, AEL can read it.
The headline numbers come from real sessions and parametric simulation. We're transparent about the assumptions — read the full methodology on the Research page. Real deployments typically see 2–4× improvement depending on session length, tool usage, and context accumulation rate.
V_shed is the cost of irrelevant tokens currently in your agent's context — what you're paying every turn to re-read content that's no longer useful. When V_shed exceeds S_real (the cost to spawn fresh), spawning saves money.
AEL (Agent Efficiency Limit) is the break-even point — the first moment where resetting a session costs less than continuing it. MEL (Maximum Efficiency Limit) is the optimal point — the spawn timing that maximizes total savings over the full session, derived in closed form from your session’s actual economics. AEL tells you when spawning becomes profitable. MEL tells you when it’s most profitable. The dashboard surfaces both signals live.
Every session lives in one of three zones. Zone 1 (before AEL): resetting costs more than it saves. Zone 2 (AEL to MEL): resetting is profitable and savings are growing toward the peak. Zone 3 (past MEL): savings are still positive but declining with every additional turn. The SpawnPoint Dashboard tells you which zone you’re in at every turn, and lets you set a personal waste tolerance threshold for when you want to be notified.
That’s the right question, and we appreciate you asking it directly. The MEL derivation — including the closed-form expression for the optimal spawn interval N* and the threshold formula — is covered under provisional patent applications filed with the USPTO and is not being disclosed publicly at this time.
What we will say: it’s a clean calculus result derived from minimizing total session cost as a function of spawn interval. It produces a single number, computable in real time from your session’s own data. No tuning, no configuration — it falls out of the math.
If you independently discover a spawn timing method that consistently outperforms MEL on total session cost — validated across session types and lengths — we’d genuinely want to hear about it. Reach out. We’re not worried about the competition. We’re confident in the derivation.
Yes. We ran identical simulations across Anthropic Claude Sonnet and OpenAI GPT-5.4 — 9 session type and length combinations on each, 18 scenarios total. MEL produced the highest net savings in every scenario on both providers, with zero exceptions. The framework is parameterized by three pricing inputs and derives correct thresholds for any provider automatically. Full results on the Research page.
No. One-time purchase. No monthly fees, no recurring billing. You pay once and keep the software, including free updates.
At $4.99, the Dashboard can pay for itself very quickly. If you’re running a mid-length agent session (50–100 turns) with Anthropic or OpenAI, context drag alone can cost a few dollars in wasted cache reads. Catching even one unnecessary continuation can cover the purchase price. After that, every session it catches is pure savings.
Enterprise licensing (per-seat, usage-based, or annual contract) is available for teams running multiple agents. Get in touch to discuss.