Which clients can use the Alma MCP server?

Any MCP-compatible client — Claude Desktop, Cursor, Windsurf, Claude Code, Continue. Same package, same config shape; only the config file location differs per client.

What tools does it expose?

35 tools across 4 categories: memory (alma_remember, alma_recall, alma_search, alma_extract), context (alma_assemble, alma_continue, alma_focus), Soul (alma_update_block, alma_get_block) and code workspace (8 tools: list/read/write files, search, skills, apply diff).

On which plan does the MCP server work?

Every paid plan (Starter, Pro and Max). Your Alma plan caps how many memories you can store and your monthly budget; the MCP server itself has no extra restrictions.

How is authentication handled?

Via your Alma API key (Settings → API Keys). The key goes in the MCP client config under env (e.g. ALMA_API_KEY ). Each client process holds the key locally — never sent anywhere except api.alma.olivares.ai.

MCP Server

The Alma MCP server (@olivaresai/alma-mcp) connects Alma's persistent memory to any MCP-compatible AI client — Claude Desktop, Cursor, Windsurf, Continue, Claude Code. 35 tools (memory CRUD, search, context assembly, Soul blocks, code workspace) plus 10 resources. Install with npm install -g @olivaresai/alma-mcp and add to your client config — total setup ≈ 5 minutes.

Model Context Protocol server with 35 tools and 10 resources. Connect Alma to any MCP-compatible AI tool — Claude Desktop, Cursor, Windsurf, Continue. Install via npm: @olivaresai/alma-mcp. Provides native tool calls for memory search, context assembly, block updates, and more.

See plans