# memory engine — agents.txt
# https://docs.memory.build/agents.txt
#
# This file describes how AI agents should interact with memory engine.

name: memory engine
description: Permanent memory for AI agents. Store, search, and organize knowledge across conversations.
url: https://memory.build
docs: https://docs.memory.build

# MCP server
mcp:
  install:
    opencode: me opencode install
    codex_cli: me codex install
    claude_code: me claude install --mcp-only
    claude_code_plugin: me claude install
  description: Memory engine ships as an MCP server. OpenCode, Codex CLI, and Claude Code all support MCP-only install via `me <agent> install` (Claude Code: `me claude install --mcp-only`). For Claude Code, the default `me claude install` installs the full plugin (hooks + slash commands + MCP) via Claude Code's plugin marketplace.
  compatible_clients:
    - Claude Code
    - Codex CLI
    - OpenCode
    - Any MCP-compatible client

# CLI
cli:
  install: see https://docs.memory.build/getting-started/
  binary: me
  description: Full-featured command line interface. Store, search, and manage context from your terminal.

# TypeScript client
sdk:
  install: npm i @memory.build/client
  description: Programmatic access for agents and applications.

# Search capabilities
search:
  - keyword: BM25 via pg_textsearch
  - semantic: HNSW via pgvector
  - hybrid: RRF (Reciprocal Rank Fusion)
  - faceted: jsonb + GIN index
  - temporal: tstzrange + GiST index
  - hierarchical: ltree + GiST index
  - grep: POSIX regex filter on content