Research notes
Benchmark methodology, engineering decisions, and observations on building persistent memory infrastructure for multi-agent AI systems.
Five things Iranti solves that memory layers don't
Most AI memory tools store facts and retrieve them by similarity. That solves about 20% of the problem. The other 80% is coordination: agents sharing state across tools, work surviving crashes, conflicting beliefs getting resolved, teams knowing what their agents actually know.
6,000 downloads in three weeks: what we learned building agent memory infrastructure
Three weeks after publishing to npm, Iranti had 6,000 downloads with no Product Hunt launch. The adoption pattern -- who installed it, how they were already using agents, which failure modes they were hitting -- tells us more about the current state of multi-agent tooling than the number does.
GitHub Copilot CLI now supports Iranti shared memory
As of v0.3.17, iranti copilot-setup wires GitHub Copilot CLI into the same shared project memory as Claude Code and Codex. How the integration works, what's different about Copilot's hook system, and what it means when all three tools share one memory store.
How to share context between Claude Code and Codex (without re-briefing)
Every time you switch from Claude Code to Codex, it starts blank. Here's the problem, the current workarounds developers are using, and the one setup that actually solves it.
Iranti: a persistent memory MCP server for AI agents
Iranti ships a stdio MCP server that any MCP-compatible client can connect to. Connect Claude Code, GitHub Copilot, Codex, or your own agent and get structured, persistent, cross-session memory with exact retrieval, conflict handling, and operator visibility.
Iranti vs Mem0: what the benchmarks actually show
A direct comparison across four benchmarks: recall accuracy, pool efficiency, conflict resolution, and cross-session persistence. Where each system wins and where the architectural tradeoffs land.
How to give Claude Code persistent memory across sessions
Claude Code starts every session with no memory of previous work. Iranti adds a persistent MCP memory layer in one command. How it works, what it stores, and what changes in practice.
Your AI research assistant shouldn't lose its memory every session
Three research workflows where persistent agent memory eliminates the most frustrating part of working with AI: literature review that builds across sessions, hypothesis tracking that survives experiment cycles, and manuscript writing with real continuity.
Why Iranti uses 37% fewer tokens in long coding sessions
We measured cumulative input token usage over a 15-turn coding session with and without Iranti. By turn 15, the Iranti arm uses 37% fewer tokens. Here's exactly how we measured it and why the gap grows over time.