MyAITwin MCP

The toolbox.A production-grade RAG database with semantic search that you shape from chat. You define the structure, the types, the tags. It is yours, it is visible, and you are the architect of it.

Or use the canonical OAuth-authenticated endpoint:

- URL:https://myaitwin.lutolearn.com/mcp
- Transport:Streamable HTTP
- Auth:OAuth 2.1 with PKCE (S256) and Dynamic Client Registration

Requires a client with MCP capability. Currently Claude Pro, Claude Team, and ChatGPT Pro.

All tools are annotated withtitle,readOnlyHint, anddestructiveHintper the MCP spec. Of the 19: 10 read-only, 8 write (non-destructive), 1 destructive (delete_knowledge).

RAG is Retrieval-Augmented Generation. It is the architecture that lets AI answer using your specific knowledge rather than its training data alone.

- Supabase (PostgreSQL)for structured records with types, tags, and provenance.
- Pineconefor vector embeddings, so you can search by meaning rather than exact words.

When you search, both layers work together and return results ranked by relevance. Every result is cited with source and date, and tagged with provenance: personal (your own thinking), organisational (from your organisation), or external (from someone else).

The architectural insight worth getting right:

Knowledgeis what you know. Facts, decisions, transcripts, observations.

Skillsare how you express things. Your LinkedIn voice. Your email style. Your proposal structure.

Exceptional output needs both. Take a meeting transcript and ask for a follow-up email. The twin needs the transcript and your email skill to produce something that is accurate and unmistakably yours. Neither alone is enough.

- Bearer token authentication on every request, hashed at rest.
- OAuth 2.1 with PKCE for connector-style integration. No shared secrets, no static credentials.
- Multi-tenant data isolation: each user lives in their own namespace. Other users can never read your data. Verified by a 35-check cross-tenant test suite.
- Rate limiting per tenant.
- Append-only audit log on every tool call.
- Prompt injection guardrails on stored content.
- Your data is used only to provide the service. Never used to train AI models. Never shared with third parties.
- You can delete your account and all data instantly from/create. Deletion is immediate and irreversible.

Local-first agent memory: a plain-Markdown Obsidian vault is the source of truth, with a rebuildable DuckDB index for hybrid BM25 + vector + graph recall.

Local-first second brain for Claude. Saves notes as Markdown and recalls them across sessions with offline semantic search, no cloud and no API keys.

Local-first MCP memory server with no external dependency, source citations and OKF/Markdown KB.

Open-source shared AI memory and team knowledge layer with 44 MCP tools for notes, memories, URLs, emails, projects, graph search, semantic search, and Git sync.

Self-maintaining knowledge vault: figure-level search, auto-wikilinks, and sleep-based memory compression.

Local Work Model for AI agents that learns from real outcomes.

Auditable, self-improving knowledge & memory for AI agents over MCP — citation-enforced answers and a replayable why-trace, self-hosted on Postgres.

Agents Remember is a Drift-aware repository memory for coding agents in complex codebases. Captures what code can't say on its own! Retrieves memory by path, semantic search, and relationship (code-graph).

Self-hosted MCP server giving AI agents persistent memory — Markdown source of truth, hybrid BM25+embedding search, typed graph relations.