If you have not cloned the GitHub repo Ponytail, you absolutely need to. The token efficiency is off the charts, especially if you're running Fable...
If you have not cloned the GitHub repo Ponytail, you absolutely need to. The token efficiency is off the charts, especially if you're running Fable 5. I did a migration today for a client that was previously running a headless dash P stateless agent to a full stateful agent with memory structure.
In about 4 hours on extra high without hitting limits. We did a full teardown of the current agent, full rebuild of the new agent, redid the memory architecture. We added Google Workspace CLI, daily briefings, boot contacts compiler, four different lanes for trackers, a watchdog demon for self-healing, neural memory, graphify, an automatic handoff and reboot at like 3:40 in the morning, and a bunch of other stuff.
Ran for 4 straight hours on extra high and Fable 5 didn't hit the session limits. I've done this before with like Opus 4.8 and it lasted maybe 2 and 1/2 3 hours including testing before it hit session limits. So, absolutely incredible.
Ponytail GitHub, be sure you get it.
See this running in your business — free 30-minute consultation