Provider Setup¶

memoire supports six AI assistant providers. Each gets the same full causal graph and MCP tools — what differs is the instructions file format, MCP config location, and LLM used for markdown extraction.

Overview¶

Provider	Instructions file	MCP config	Activity hooks	Markdown LLM
Claude Code	`CLAUDE.md`	`.claude/settings.json`	✓ PostToolUse / PreToolUse	`claude --print` CLI
Cursor	`.cursor/rules/memoire.mdc`	`.cursor/mcp.json`	—	Anthropic API
Windsurf	`.windsurfrules`	`~/.codeium/windsurf/mcp_config.json`	—	Anthropic API
Codex CLI	`AGENTS.md`	`.codex/config.toml`	—	OpenAI API
Gemini CLI	`GEMINI.md`	`.gemini/settings.json`	—	Google API
Ollama	—	—	—	Local Ollama

Claude Code¶

The most fully integrated provider. Claude Code's hook system lets memoire track every file read and edit, enabling temporal causality inference (sequential edits → inferred DRIVES edges).

memoire init --provider claude

What gets created:

CLAUDE.md — instructs Claude to call get_context at session start and expand() before reading files
.claude/settings.json — PostToolUse hook (captures all tool calls), PreToolUse hook on Read (nudges expand first), MCP server registration

Markdown extraction: uses the claude --print CLI — no API key required, uses Claude Code's existing authentication.

Cursor¶

export ANTHROPIC_API_KEY=sk-ant-...
memoire init --provider cursor

What gets created:

.cursor/rules/memoire.mdc — MDC format with alwaysApply: true frontmatter, same instructions as CLAUDE.md
.cursor/mcp.json — registers the memoire MCP server

Markdown extraction: Anthropic Messages API. Requires ANTHROPIC_API_KEY.

Note

Cursor does not currently expose a general hook system equivalent to Claude Code's PostToolUse. The filesystem watcher handles file change tracking. Temporal causality inference (sequential edit patterns) is not available.

Windsurf¶

export ANTHROPIC_API_KEY=sk-ant-...
memoire init --provider windsurf

What gets created:

.windsurfrules — project-level rules file, same instructions content
~/.codeium/windsurf/mcp_config.json — global MCP server registration

Markdown extraction: Anthropic Messages API. Requires ANTHROPIC_API_KEY.

OpenAI Codex CLI¶

export OPENAI_API_KEY=sk-...
memoire init --provider codex

What gets created:

AGENTS.md — Codex CLI's project instructions file
.codex/config.toml — MCP server registration

Markdown extraction: OpenAI Chat Completions API (gpt-4o-mini by default). Requires OPENAI_API_KEY.

Override the model:

memoire init --provider codex --model gpt-4o

Google Gemini CLI¶

export GEMINI_API_KEY=...
memoire init --provider gemini

What gets created:

GEMINI.md — Gemini CLI's project instructions file
.gemini/settings.json — MCP server registration

Markdown extraction: Google Generative Language API (gemini-1.5-flash by default). Requires GEMINI_API_KEY.

Override the model:

memoire init --provider gemini --model gemini-1.5-pro

Ollama¶

memoire init --provider ollama

Ollama is a local inference server, not an IDE assistant — it has no standard project instructions file or hook system. memoire uses it purely for markdown LLM extraction.

What gets created: nothing IDE-side. The filesystem watcher tracks all file changes.

Markdown extraction: local Ollama at http://localhost:11434 (llama3 by default). No API key needed.

Override the model or URL:

memoire init --provider ollama --model mistral

To use a non-default Ollama URL, set it in .memory/config.json:

{
  "provider": "ollama",
  "llm": "ollama",
  "llm_model": "mistral",
  "ollama_url": "http://192.168.1.10:11434"
}

Switching providers¶

Re-run memoire init --provider <new> at any time. The config in .memory/config.json is updated and the new provider's files are created. Existing files from the previous provider are not removed automatically.