03 / DEVELOPER & CLI PLUGINS

Your AI knows your project.

Memwyre plugins for Claude Code, Grok Bot, OpenClaw, and OpenCode. Automatic context injection on session start and automated transcript capture on exit.

Get Developer Plugins → View Documentation

Session Hooks — claude / openclawLive Hook

$ claude --plugin @memwyre/claude-memwyre

[SessionStart] Project scope detected:

✔Injected 18 past memories for /workspace/src

✔Conventions active: Strict TypeScript, Tailwind v4

✔Stop hook armed: auto-syncs session transcript

// Continuous autonomous memory across agent runs

>Claude Code: "Applied architecture pattern from session #14."

The Problem

Every AI tool starts from zero.

01 . PROBLEM

No Persistent Project Memory

Command-line agents only see what you paste into the immediate prompt. When the terminal session exits, all context is lost.

02 . PROBLEM

Repeated Explanations

Engineers spend precious time repeatedly explaining dependency versions, build prerequisites, and architectural standards.

03 . PROBLEM

Multi-Agent Blindspots

When running multiple bots (e.g. Researcher, Coder, Tester in Grok or OpenClaw), agents cannot see what their peers have solved.

Plugin Specs

Install once. Remember forever.

Deterministic Shell Hooks

SessionStart injects relevant memory based on workspace root. Stop captures terminal transcript and posts it to extraction backend.

Non-Blocking Execution

Hooks are strictly non-blocking (always exit 0) and silently fail open if the network is unreachable, never halting your terminal.

Curation & Inbox Control

Captured insights are placed in your Memwyre Inbox where you can review, approve, or discard memories before indexing.

~/.openclaw/config.json

v2.0.14

{
  "plugins": {
    "entries": {
      "openclaw-plugin": {
        "enabled": true,
        "config": {
          "apiKey": "bv_sk_your_api_key_here",
          "hostUrl": "https://server.memwyre.tech"
        }
      }
    }
  }
}

Documented Integrations

Built for the tools you love.

Claude Code

@memwyre/claude-memwyre lifecycle hooks (SessionStart & Stop transcript archival).

Package Active

Grok Bot

@memwyre/grok-plugin cross-bot shared memory across Researcher, Coder, and Ops.

Zero Dependencies

OpenClaw

@memwyre/openclaw-plugin persistent memory provider for autonomous agents.

v2.0.14 Ready

OpenCode

memwyre-opencode-plugin auto-injecting shared conventions and architecture patterns.

Marketplace v1.1.5

Technical Specifications

Technical specifications of our developer plugins.

Memwyre developer plugins integrate directly into command-line and agent execution cycles. For CLI tools like Claude Code, the package (@memwyre/claude-memwyre) exposes two primary lifecycle hooks in hooks.json: SessionStart which executes inject-memory.cjs to retrieve workspace memories matching your project directory, and Stop which executes capture-session.cjs to parse and upload your session transcript upon exit.

For autonomous frameworks like OpenClaw, the plugin (@memwyre/openclaw-plugin@2.0.14) implements a persistent memory provider. When agent tool profiles are configured to full or coding, OpenClaw gains direct access to save_memory and search_memwyre, enabling continuous cross-run recall.

In multi-bot environments like Grok Bot, the plugin (@memwyre/grok-plugin) enables cross-bot shared memory. Built with zero external npm dependencies, it ensures that findings discovered by your Researcher bot are instantly available when your Coder or Ops bot initializes, eliminating duplicate work across your roster.

For OpenCode, the plugin injects shared architecture decisions and conventions via get_conventions and search_memwyre, preventing stylistic deviations and ensuring generated code complies with your team's established standards.