Give agents persistent semantic memory with Memora
Use Memora to give MCP-compatible agents persistent semantic memory, document recall, and knowledge-graph context across sessions.
npx skills add agentskillexchange/skills --skill give-agents-persistent-semantic-memory-with-memora
Use Memora when an agent needs a durable memory layer that survives across sessions and can store, search, link, deduplicate, and retrieve facts, documents, TODOs, issues, and conversational context. The operator installs the Memora server, configures an MCP client such as Claude Code, Codex, Cursor, or another MCP host, chooses local SQLite or cloud-backed storage, and keeps memory operations reviewable through the agent’s MCP tool calls.
What this skill actually does
This is skill-shaped because the repeatable job is narrow: provision a persistent semantic-memory MCP server, connect it to a specific agent runtime, define storage boundaries, then use explicit memory search/create/update/delete operations during future work. It is not a generic cache library, vector database, RAG framework, or memory product listing; invoke it when cross-session agent memory needs a bounded operational store rather than prompt-stuffed context.
Inputs and prerequisites: Python package installed from the Memora GitHub repository; MCP-compatible client such as Claude Code, Codex, Cursor, or another MCP host; optional SQLite, S3/R2, or Cloudflare D1 storage.
Setup notes: Install with pip install git+https://github.com/agentic-box/memora.git, then run memora-server for stdio MCP mode or configure memora-server –transport streamable-http for HTTP mode. Add the documented memora MCP server block to the target agent client and set storage environment variables such as MEMORA_DB_PATH or MEMORA_STORAGE_URI before using memory tools.
Source and verification boundary: use https://github.com/agentic-box/memora as the canonical reference before running the workflow; keep commands, API calls, CLI usage, and generated outputs reviewable against that upstream source.
Framework fit: publish this as a MCP workflow only when the operator can invoke the documented toolchain directly, rather than treating the upstream project as a generic product listing.