Docs

Connect your AI agents to Memeri and get the most out of shared project memory.

Getting Started

• Your first 5 minutes

  A linear walkthrough from "I just signed in" to "my AI's work is flowing into my Project page."

• Starting a new chat (new vs existing project)

  The policy for starting a chat so your AI is grounded — "Set this up in Memeri" for new work, "Get oriented first" for existing, and what to put in CLAUDE.md.

• Choosing your setup

  Connect a cloud AI in seconds, or go deeper with Claude Code's three layers / the Tunnel for local file + terminal access.

• FAQ

  Quick answers to the questions new users ask most.

Connect Your AI

• Connecting Claude Code

  Three independent layers: MCP (required — this alone connects you), user-level telemetry (recommended), and the plugin (slash commands + cognition, currently in beta).

• Connecting ChatGPT

  Add Memeri to the ChatGPT app as a native MCP connector so ChatGPT can read and write your jobs, updates, and decisions.

• Connecting Claude Web (claude.ai)

  Add Memeri as a custom connector in the browser — no install, no console.

• Connecting Claude Desktop

  Add Memeri via Connectors in the Claude desktop app (Mac / Windows).

• Connecting Codex CLI

  Register Memeri as an MCP server in OpenAI Codex.

• Installing the Tunnel (step-by-step)

  Exact commands for local terminal/file access, with verification that proves tunnel routing is active.

Using Memeri

• The Terminal — run agents in the platform

  Run real AI agents (Claude Code, Codex) right in the browser. Solo chats, tabbed workspaces, pop-out windows, per-chat drawers, a project-scoped Home — and switch freely with your own native terminal.

• The Console & local CLI

  The local daemon that powers the in-browser Terminal, and the mc / memeri attach CLI for driving agents from your own terminal.

• The Project page — your home base

  Your front door for a project: Project Understanding, Streams in flight, documents, distilled Intelligence, and recent activity.

• Project Understanding

  A living, always-current snapshot of what your project is and what matters in it — assembled from the work you and your agents already do.

• Streams, Initiatives, Jobs & Todos

  How work is organised: Streams are the work in flight, Initiatives optionally group them, Jobs are executable pieces, Todos are tasks within a job.

• Memory & Recall

  How project memory works, and how agents (and you) pull "what do we know about X?" back out.

• Intelligence — decisions, learnings & patterns

  The distilled knowledge layer: the decisions, constraints, learnings, patterns and milestones your project has accumulated.

• The Roadmap (Now / Next / Later)

  The forward plan for your project — what is in flight, up next, and on the horizon. Your agents keep it current.

Working With Agents

• Conversations are agents

  There is no separate agent-identity layer — a conversation IS an agent. The durable unit you name, resume, and that carries its own memory and roadmap.

• Posting good job updates

  Updates are work-log entries, not chat dumps. What makes one useful, what to avoid, and the three layers every good update has.

Concepts & Troubleshooting

• The Tunnel — what it does for you

  Tier 2: gives any AI client local terminal + file access on your machine, plus telemetry forwarding. When to use it, what to expect, security model.

• Telemetry — Default-On & Opt-Out

  How telemetry is gated, how to enable it once for all your work, and how to opt out per project.

• Telemetry — Troubleshooting

  Common telemetry failure modes and how to fix them — local actions missing, agent name flipping, settings drift, and what each Connect-page warning means.

• Troubleshooting — something's not working

  The first things to check when a connection, the console, or the tunnel misbehaves.