What We Shipped in June 2026: Completing the AI Coding Tool Memory Series

June 2026

June closed out a milestone we've been building toward for months: every major AI coding tool now has a dedicated MemNexus guide. That's 21 tools, one persistent memory layer, and a reference library that didn't exist six months ago. Here's what shipped, what we fixed, and what we're seeing in the data.

The complete AI coding tool memory series — 21 tools covered

In June we published the final five guides in the series: Devin, Gemini Code Assist, Augment Code, Supermaven, and ChatGPT. That completes coverage of the AI coding tool landscape as it stands today.

The series now spans the full range of how developers actually use AI — from inline completion tools like Supermaven and GitHub Copilot, to autonomous coding agents like Devin, to general-purpose assistants like ChatGPT. Each guide follows the same structure: what resets between sessions, what the tool's built-in memory mechanisms can and can't do, and how to connect MemNexus to get persistent memory that works across all of them.

The practical payoff is that no matter which tool you use, there's a guide written for your specific setup. You're not adapting a generic tutorial. The setup steps, the example commands, and the caveats are specific to how that tool handles MCP.

The deeper value is what the series signals: MemNexus is tool-agnostic by design. The memory layer sits outside any single AI tool. Your decisions, debugging history, and project conventions follow you whether you're in Cursor, Devin, or Claude Code — and across all of them simultaneously.

Try it yourself. Sign up free and run mx setup to connect your tools.

build_context: one call to get up to speed

build_context is now the feature we lead with on the homepage, and the response from developers has confirmed why.

The idea is simple: before your agent starts work, it calls build_context and gets back a structured briefing — active work, key facts, gotchas, recent activity — in around 200ms. Instead of asking what you were working on, your agent already knows. Instead of rediscovering a documented bug, it sees the gotcha surfaced proactively.

What makes it sticky is the gotcha detection. You don't tag memories as gotchas. The system identifies facts that appear across multiple separate memories — which means they've come up enough times that they're worth flagging at session start. That pattern detection is emergent from your save history, not from any manual curation.

We published a dedicated tutorial for build_context this month alongside the homepage update. If you haven't tried it yet, the tutorial walks through exactly what you get back and how to hand it to your agent at session start.

The goal for the next release is automatic delivery via startup hooks — so the briefing arrives the moment a new session begins, without the agent needing to call the tool at all.

The things we fixed

The package name was wrong. For two months.

This is a build-in-public post, so: we had the wrong package name in our install command across the entire site for roughly two months.

The customer CLI is @memnexus-ai/cli. That's the package that installs the mx binary. For two months, a different package name appeared in install commands on the homepage, in 42 blog posts, and across 12 docs pages — 61 files in total.

npm install -g @memnexus-ai/cli

PR #3497 fixed all 61 files in one pass. Anyone who hit the old install command and got an error or the wrong package — that's on us. The correct command is above and is now consistent everywhere.

Analytics: cross-subdomain tracking and cleaner data

We fixed PostHog tracking across subdomains so that a session starting on memnexus.ai and continuing to portal.memnexus.ai is correctly recorded as a single session rather than two separate ones. Previously, cross-subdomain navigation was resetting session identity.

We also added production-only filtering to our analytics pipeline. Dev and test traffic was mixed into our reported numbers, which made it harder to read what real users were actually doing. That's now filtered out at ingestion. The data we're acting on reflects production behavior only.

Finally, we fixed CTA tracking to record accurate destination URLs. Clicks were being tracked but the destination was being lost in some cases. That's corrected — we can now see which CTAs are converting and where they're sending people.

What the traffic data tells us

Blog content now drives 44% of all site traffic. That's the clearest signal we have that the "[tool] persistent memory" keyword pattern is the right content strategy.

The top performers are the Codex, Cursor, and GitHub Copilot posts. These are developers searching for a specific answer to a specific problem — "how do I make this tool remember my project" — and finding a guide that answers it directly. The branded long-tail pattern works because it matches the actual search intent.

Two things happened in June that we hadn't seen before. The /vs/letta comparison page entered Google search results for the first time. And claude.ai and copilot.microsoft.com appeared as referral sources — users clicking from AI conversations to our site. That's a new traffic channel that didn't exist in earlier months: people asking AI assistants about persistent memory for AI coding tools, and those assistants surfacing MemNexus.

We're not projecting where that goes. But it's notable that it appeared at all, and that it appeared alongside completing the 21-guide series.

What's next

The series is complete, but the content work isn't done. We're publishing more framework and workflow integration guides — the same depth as the tool series, applied to specific development patterns. Think: how memory changes your workflow when you're building with a particular agentic framework, or how to structure memories for a specific kind of project.

On the product side: startup hooks for automatic build_context delivery, continued docs improvements, and better onboarding for developers who are new to MCP-based memory.

Get started

If you've been meaning to try MemNexus, the full series being complete is a good moment. Find the guide for the tool you already use, follow the setup steps, and run mx setup to connect it.

npm install -g @memnexus-ai/cli
mx auth login
mx setup

Sign up free at portal.memnexus.ai

---

Explore the full series

Every major AI coding tool, covered:

• Cursor · GitHub Copilot · Claude Code · Windsurf · Codex · Devin · ChatGPT
• Full list of all 21 guides →