wikimemory
Personal, remotely hosted memory service for Claude, Codex, and other MCP clients that preserves research, project state, and decisions in an auditable revision store with a web UI.
README
Wikimemory
Wikimemory is a personal, remotely hosted memory service for Claude, Codex, and other MCP clients. It preserves research, project state, technical decisions, and reusable context in an auditable revision store that the owner can browse and search from the web.
It includes remote HTTP MCP with OAuth/PKCE, passkey identity, append-only revision storage, full-text search, a React browse/search/history UI, multi-passkey controls, sanitized exports, and version-matched client skills. The server does not call an LLM; Claude or Codex performs synthesis while Wikimemory provides deterministic, auditable storage and retrieval.
What you can use it for
- Maintain a to-do list and let an assistant help prioritize, update, and archive it.
- Track personal or work projects with status, decisions, open questions, and next actions.
- Collect sources and build a durable research notebook with summaries and provenance.
- Keep an automated journal or work log containing the entries you choose to save.
- Maintain notes about people, conversations, commitments, and important context.
- Preserve meeting notes and decision history so later sessions know why a choice was made.
- Build a personal assistant that combines connected email or calendar tools with selected long-term memory.
- Keep learning notes, reading lists, household context, or a troubleshooting and incident notebook.
Wikimemory does not read email, calendars, or other services on its own. A connected assistant uses those tools and saves only the information you ask it to remember. Because these uses can involve sensitive personal data, review the security model before deciding what to store.
Quick start
Requirements: Node.js 22+, npm, a Cloudflare account, and a passkey-capable browser or password manager. No Google Cloud project or OAuth application is required.
npx wrangler login
npx wikimemory install
The installer previews the exact Cloudflare account, Worker, D1 database
(Cloudflare's managed SQLite), and OAuth/session store before creating anything.
Open its one-time URL to register the owner passkey. The installer also prints the
exact /mcp URL to paste into clients that require manual connector setup. Then
connect a command-line client and install its memory skills:
npx wikimemory connect codex
npx wikimemory skills install codex
# or
npx wikimemory connect claude
npx wikimemory skills install claude
To use Wikimemory on a phone, first add the HTTPS /mcp URL printed by the installer
or by npx wikimemory status:
Use the complete URL, including /mcp. The deployment root opens the Wikimemory web
application and is not an MCP endpoint.
- Claude: add it as a custom connector in Claude's web settings. It then appears in Claude mobile.
- ChatGPT: enable developer mode on ChatGPT web, create an app using the
/mcpURL under Settings → Plugins, and finish passkey authorization. The app then appears in ChatGPT mobile.
See the complete installation guide for detailed connection steps, recovery, passkey management, upgrades, and safe uninstall.
Run locally
npx wikimemory dev
This starts the Worker, React app, D1, and KV through local Wrangler emulation.
State persists under ./.wikimemory/dev; production passkey identity is
replaced with a clearly marked fake local owner while OAuth, PKCE, scopes, tokens,
and MCP transport remain real.
Lifecycle commands
npx wikimemory status
npx wikimemory browse
npx wikimemory upgrade
npx wikimemory recover
npx wikimemory passkeys list
npx wikimemory uninstall # preview only
Parallel installations use --deployment NAME. Non-secret lifecycle state lives
under ~/.config/wikimemory/deployments/NAME/. Uninstall requires an explicit apply
step and exact Worker-name confirmation because deleting D1 permanently destroys the
stored memory.
How it works
Claude / Codex ---- Streamable HTTP MCP ----+
|
Browser ------------ HTTPS web app ---------+--> Cloudflare Worker --> D1
|
Passkey ------------ owner authentication --+
Cloudflare is the recommended host because Workers, D1, and KV all have free tier allowances large enough for typical personal use; the paid Workers plan starts at $5/month if those limits are exceeded. Memory is protected by access control and Cloudflare-managed encryption, but it is not end-to-end encrypted from Cloudflare. Do not store credentials or material you are unwilling to entrust to the host.
Design documents
- Product specification
- Architecture
- Security and threat model
- Data model
- MCP contract
- Local development and testing
- Installation and client connection
- Export formats and privacy
- Pasteable agent instructions
- Implementation history
- Independent design review and dispositions
Current scope
Wikimemory supports one owner and one workspace, authenticated with passkeys. It includes a React web application with owner administration, remote MCP, local emulation, portable export, agent skills, and guided installation and client setup.
Multi-tenant SaaS, vector search, attachments, built-in chat, offline operation, manual document editing, scheduled backups, monitoring infrastructure, custom domains, and permanent staging infrastructure are outside the current scope.
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.