Tephra MCP

Documentation and config examples for Tephra, a hosted remote MCP server: one memory, tied to your account, shared by every AI you connect. This repository contains no source code; Tephra is a paid hosted service by Obsidian Metrics.

• Server URL (Streamable HTTP): https://tephra.obsidian-metrics.com/mcp

• Connecting: add the server URL. Clients that support MCP OAuth can sign in from the client; for others, use a personal AI key (tpk_...). Sign-in support varies by client, so check the connect guide for yours.

• Connect guide: CONNECT.md and https://tephra.obsidian-metrics.com/connect

• Registry metadata: server.json; client config examples: mcp-config-examples.json

Features

• One memory for every AI you connect. Claude, Cursor, Claude Code, VS Code, Gemini CLI, Windsurf, Zed, Le Chat, Perplexity and other MCP apps all read and write the same memory.

• Three MCP tools: tephra_recall (ranked recall over your memory plus the shared library, query up to 2000 chars, limit 1-25), tephra_search (semantic search, limit 1-25), tephra_remember (write a memory to your own account: text up to 8000 chars, optional region, up to 10 tags, importance 1-5).

• Hybrid recall: bge-m3 semantic search + BM25 keyword search fused by reciprocal rank, cross-encoder rerank, and a graph walk over linked memories; 2.5 s recall budget.

• Account-scoped by default: each member is their own tenant; everything a member saves is tied to their account and is not visible to other members.

• Shared library: public knowledge (no member data), grown hourly from sources such as arXiv, PubMed, Federal Register, SEC EDGAR 8-K filings and Crossref/OpenAlex; recall draws on it alongside your own notes.

• Control and audit: per-key revocation, read-only keys, key activity log, edit history, edit/delete/contest a memory, data export (app).

• Prompt-injection hardening on key/MCP results: invisible characters stripped, records capped at 4000 chars, results wrapped as notes-not-instructions.

• Remote streamable-HTTP MCP, hosted on Cloudflare Workers; nothing to install or self-host.

Related MCP server: jalipata-mcp-memory

Limits

• Tephra is a paid product: no free tier. Access is Premium ($19/mo via the Skool community, $29/mo direct) or a $399 one-time Tephra seat; no usage metering.

• The MCP tools return ranked records only; the "you may be missing" completeness alert, summary sheet and push directive are not passed through MCP tools.

• The AI must call tephra_remember; Tephra does not auto-capture chats and has no Gmail/Slack/Drive connectors.

• No image/audio ingest. Members cannot publish into the shared library.

Benchmark

LongMemEval-S, all 500 questions, top-10 retrieval, gpt-4o-2024-08-06 judge, one run: 85.2% with a gpt-4o reader; 95.2% with a Claude Opus 5.5 reader (run via Claude Code sub-agents); same retrieval. Retrieval had at least one needed session in the top 10 for 467 of 470, and all needed sessions for 435. Not a ranking.