AI agents die mid-task when context fills — losing every decision and WHY it was made. agent-recall-ai fixes this with Decision Anchors: critical reasoning is protected from compression and never pruned. Zero-code for Claude Code: pip install agent-recall-ai && agent-recall-ai install-hooks Works with OpenAI, Anthropic, LangChain, LangGraph, CrewAI. Includes live dashboard, PII redaction, and cost monitors.
No reviews yetBe the first to leave a review for agent-recall-ai
Maker
📌
Hey Product Hunt! 👋 I'm Srinath, and I built agent-recall-ai.
The backstory: I was doing a complex auth refactor with Claude Code. 2 hours in, the context window filled and the session died. Every decision, every constraint, every "why" — gone. Starting from scratch felt wrong. So I built structured auto-save for AI agents.
The key insight is Decision Anchors — messages containing words like "decided", "rejected", "because", "constraint" are mathematically protected from context compression. The reasoning chain survives forever, no matter how long the session runs.
Zero-code setup for Claude Code users:
pip install agent-recall-ai
agent-recall-ai install-hooks
Two commands. Every session protected from that point on.
Would love to hear from anyone building long-running agents — what's your biggest pain point when sessions die? Happy to answer any questions about the architecture or design decisions!