I built a persistent memory system for AI agents.
AgentRecall — your agent remembers everything across sessions. Decisions, corrections, mistakes, context. All local markdown, zero cloud, zero telemetry. the more complicated your project is, the more tokens can be saved.
Two commands that change everything:
/arstatus — one command, full picture. Every project you're working on, what's pending, what's blocked, where you left off. Walk into any session and know exactly what to do.
/arsaveall — one command, everything saved. Journal, insights, palace rooms, awareness. No more losing context when a session ends. Your next agent starts where this one stopped.
What makes it different:
Correction-first memory — when you tell the agent "no, not that", it doesn't just fix it. It saves the correction permanently. Next session, next project, any agent gets warned: "you've been wrong about this before."
Decision trail — agent tracks prior confidence → evidence → outcome across sessions. After 3+ decisions: "your priors on API design average 0.8 but outcomes average 0.5 — you're overconfident." Metacognition for agents.
Palace rooms — persistent knowledge organized by topic with salience scoring. Architecture decisions, active goals, blockers, lessons learned. Obsidian-compatible — open it as a vault.
6 MCP tools. Works with Claude Code, Cursor, VS Code, Windsurf, Codex, any MCP client.
One line to install:
npx -y agent-recall-mcp
GitHub: https://t.co/nXznlQm5Uq
#ClaudeCode #MCP #AgentMemory #AI #OpenSource #Obsidian #DevTools #AIAgent #LLM #BuildInPublic
Introducing Claude Opus 4.8: it builds on Opus 4.7 with sharper judgment, more honesty about its own progress, and the ability to work independently for longer than its predecessors.
Available today at the same price.
I built a persistent memory system for AI agents.
AgentRecall — your agent remembers everything across sessions. Decisions, corrections, mistakes, context. All local markdown, zero cloud, zero telemetry. the more complicated your project is, the more tokens can be saved.
Two commands that change everything:
/arstatus — one command, full picture. Every project you're working on, what's pending, what's blocked, where you left off. Walk into any session and know exactly what to do.
/arsaveall — one command, everything saved. Journal, insights, palace rooms, awareness. No more losing context when a session ends. Your next agent starts where this one stopped.
What makes it different:
Correction-first memory — when you tell the agent "no, not that", it doesn't just fix it. It saves the correction permanently. Next session, next project, any agent gets warned: "you've been wrong about this before."
Decision trail — agent tracks prior confidence → evidence → outcome across sessions. After 3+ decisions: "your priors on API design average 0.8 but outcomes average 0.5 — you're overconfident." Metacognition for agents.
Palace rooms — persistent knowledge organized by topic with salience scoring. Architecture decisions, active goals, blockers, lessons learned. Obsidian-compatible — open it as a vault.
6 MCP tools. Works with Claude Code, Cursor, VS Code, Windsurf, Codex, any MCP client.
One line to install:
npx -y agent-recall-mcp
GitHub: https://t.co/nXznlQm5Uq
#ClaudeCode #MCP #AgentMemory #AI #OpenSource #Obsidian #DevTools #AIAgent #LLM #BuildInPublic