Three Slash Commands So You Never Have to Memorize MCP Tool Syntax

You have MemNexus configured. The MCP tools work. But every time you want to save a memory, you type something like this:

create_memory({ conversationId: "conv_abc123", content: "Finished refactoring the auth middleware. Switched from session cookies to JWT. Decision: stateless tokens reduce Redis dependency." })

You pause. Was the parameter conversationId or conversation_id? Do you need to pass topics, or does extraction handle that? What was the conversation ID from earlier — was it conv_abc or conv_xyz?

Meanwhile, this does the same thing:

/mx-save

Claude Code uses your current conversation context — recent tool calls, files you've touched, your git branch — to generate a memory with appropriate content, conversation ID, and context. You confirm and move on.

That is what mx setup skills gives you.

What Gets Installed

Running mx setup skills installs three slash commands into Claude Code at the global level (~/.claude/skills/mx-*/SKILL.md), so they are available in every project:

/mx-save — Quick Progress Snapshot

Use this after completing a task, fixing a bug, or making a decision. It uses your current conversation context — recent tool calls, files you've been working with, your git branch — to generate a concise memory. Conversation tracking is handled for you: the skill reuses your current conversation ID if one exists, or starts a new one.

When to reach for it:

• You just got a test passing after 30 minutes of debugging
• You made an architectural decision and want to record the rationale
• You are about to switch to a different task and want to checkpoint progress

/mx-checkpoint — Interactive Milestone Checkpoint

Use this at major boundaries: end of a feature, before a PR, after a deploy. It walks through a structured checklist — what was done, what decisions were made, what blockers exist, and what comes next. The output is more thorough than /mx-save because it prompts you for details you might skip in a quick snapshot.

When to reach for it:

• You finished a multi-day feature and want a complete record
• You are handing off work to another agent or picking it up tomorrow
• You want to document trade-offs and alternatives considered

/mx-buildcontext — Session Briefing

Use this at the start of a session. It calls build_context with your current working directory and task, then displays active work, key facts, gotchas, and recent activity. If you have read the build-context post, this is the same capability — just invoked with a slash command instead of a manual MCP tool call.

When to reach for it:

• You are starting a new session on a project you haven't touched in days
• Context compaction just wiped out earlier messages
• You want to see what gotchas have been documented before diving in

How to Install

mx setup skills

That's it. The command creates three files under ~/.claude/skills/:

~/.claude/skills/mx-save/SKILL.md
~/.claude/skills/mx-checkpoint/SKILL.md
~/.claude/skills/mx-buildcontext/SKILL.md

Restart Claude Code, and the commands appear in your slash-command menu. Type /mx- and you will see all three.

Skills are version-controlled. When you update the CLI (mx update or npm install -g @memnexus-ai/cli@latest), running mx setup skills again will update any skills whose versions have changed. If they are already current, it skips them.

# Preview what would be installed or updated
mx setup skills --dry-run

# Force reinstall even if versions match
mx setup skills --force

# See what's currently installed
mx setup skills --list

# Remove all MemNexus skills
mx setup skills --remove

A Real Session, Start to Finish

9:00 AM — Start of day. You open Claude Code on a project you haven't touched since Friday.

> /mx-buildcontext

Claude shows a briefing: you were working on the payment webhook handler, there is a gotcha about idempotency keys needing to be UUIDs, and your last session ended with retry logic untested.

9:45 AM — Tests pass.

> /mx-save

From your conversation context, Claude knows you are on feat/payment-webhooks and what test files you added. It saves a memory: "Payment webhook retry logic implemented and tested. Exponential backoff with jitter. All 7 test cases passing." Conversation tracking is handled for you.

11:30 AM — Feature complete. You are about to open a PR.

> /mx-checkpoint

Claude walks through the milestone: what shipped, decisions made, blockers, and next steps. One structured record captures all of it.

Tomorrow, when you or another agent picks this up, /mx-buildcontext will surface that checkpoint along with the gotchas from every session that touched this feature.

The Difference

The MCP tools still work. You can still call create_memory and build_context directly. Skills do not replace those tools — they orchestrate them. The difference is UX friction.

With raw MCP tools, you need to remember parameter names, construct the right content, manage conversation IDs manually, and decide what context to include. That friction means you save memories less often, which means your context store is thinner, which means build_context has less to work with.

With slash commands, the barrier drops to typing /mx-save and confirming. You save more often. Your memory store gets richer. Future sessions start better informed.

Get Started

# Install the CLI if you haven't already
npm install -g @memnexus-ai/cli@latest

# Authenticate (interactive prompt — key stays out of shell history)
mx auth login

# Set up MCP and steering rules (if not already done)
mx setup

# Install slash commands
mx setup skills

Restart Claude Code. Type /mx- and start using them.

---

mx setup skills is available in the latest CLI release. Update with mx update or npm install -g @memnexus-ai/cli@latest.

---

Want persistent memory for your AI coding agents? Join the MemNexus preview or check the docs to get started.