API Reference
Complete reference for all JauMemory APIs and tools
Introduction
JauMemory provides 50 MCP tools for persistent AI memory, automation, and integration:
- 3 Discovery Tools: search, fetch, get_guide (tool catalog + onboarding guides)
- 3 Authentication Tools: Browser-approval auth flow (mcp_login / mcp_authenticate / mcp_logout)
- 7 Core Memory Tools: remember, recall, forget, update, analyze, consolidate, memory_stats
- 8 Collection Tools: Organize memories into collections
- 7 Agent Tools: Multi-agent coordination, error learning, reflection, collaboration
- 3 Credential Vault Tools: Encrypted credential storage with rotation
- 5 Tool Registry Tools: Register reusable HTTP tools
- 3 Skill Registry Tools: Define reusable pipelines
- 2 Toolkit & Execution Tools: Semantic discovery across skills + tools, end-to-end execution
- 2 Berrry Integration Tools: Register/create Berrry-app tools
- 7 Scheduling & Task Tools: Cron schedules, task history, retriggers
Tool Naming
This reference uses snake_case tool names (mcp_login, memory_stats), as registered by the npm MCP server (@jaumemory/mcp-server). The hosted claude.ai connector exposes the same tools in camelCase (mcpLogin, memoryStats). Same tools, same parameters, different casing.
Complete Reference Available
For a comprehensive tutorial guide to all 50 tools with examples and workflows, see the Complete API Reference.
remember
MCP ToolStore a new memory with automatic classification and importance scoring.
{
content: string; // Required - The memory content
shortcuts?: string[]; // Optional - Quick flags like ["--task", "--high"]
tags?: string[]; // Optional - Tags for categorization
context?: string; // Optional - Additional context
importance?: number; // Optional - 0.0 to 1.0 (auto-calculated if not provided)
metadata?: object; // Optional - Custom metadata
}
await jaumemory.remember({
content: "User prefers dark mode with high contrast",
shortcuts: ["--insight", "--high"],
tags: ["preferences", "ui"],
importance: 0.9
});
Response
{
id: "550e8400-e29b-41d4-a716-446655440000",
type: "insight",
status: "stored",
timestamp: "2025-01-15T10:30:00Z",
importance: 0.9
}
recall
MCP ToolSearch and retrieve memories using keyword, semantic, or hybrid search. The query is optional: filters-only searches are supported.
{
query?: string; // Optional - omit for filters-only search
mode?: "keyword" | "semantic" | "hybrid"; // Default: "hybrid"
tags?: string[]; // Filter to memories with ANY of these tags
timeRange?: { // Filter by creation date
start?: string; // ISO date or YYYY-MM-DD
end?: string;
};
minImportance?: number; // Minimum importance (0.0 to 1.0)
limit?: number; // Default: 10
}
const memories = await jaumemory.recall({
query: "authentication",
mode: "semantic",
limit: 5,
minImportance: 0.7
});
get_memory (replaced by fetch)
MCP Tool
Use recall instead. The single-memory retrieval tool was removed. To retrieve specific memories, use recall with a query or filters. (The fetch tool is discovery-only: it returns tool documentation by tool ID, never memory content.) See the Complete API Reference.
analyze_patterns (folded into analyze)
MCP Tool
Use analyze instead. Pattern analysis (error→solution chains, recurring themes, learning velocity) is now part of the unified analyze tool. See the analyze section below or the Complete API Reference.
check_notifications (replaced by agent_collaboration)
MCP Tool
Use agent_collaboration instead. Multi-agent coordination is now unified under a single tool with three actions: agent_collaboration({ action: "start" }), action: "complete" (with an outcome), and action: "list". See the Complete API Reference for details.
create_agent
MCP ToolCreate specialized AI agents with personality traits and areas of expertise for multi-agent workflows.
{
name: string; // Required - Agent name
personalityTraits?: string[]; // e.g., ["curious", "analytical"]
specializations?: string[]; // e.g., ["React", "TypeScript"]
updatePrompts?: string[]; // Custom prompts
}
await jaumemory.create_agent({
name: "Frontend Expert",
personalityTraits: ["detail-oriented", "creative", "user-focused"],
specializations: ["React", "TypeScript", "CSS", "UX"]
});
save_preference (removed from MCP surface)
MCP ToolPreferences are no longer exposed via MCP. The 5 preference tools (save_preference, get_context_preferences, update_preference, delete_preference, list_preferences) were removed when preferences moved to the dashboard surface. Per-project / per-user configuration now lives in the JauMemory web app under Settings → Preferences.
backup_memories (admin-only)
MCP ToolDatabase snapshots are server-side admin operations and are no longer exposed as an MCP tool. Snapshots run on a regular schedule per the security architecture; user-facing data export is available via the dashboard's GDPR export flow.
All 50 Tools Overview
| Category | Tools | Description |
|---|---|---|
| Authentication | mcp_login, mcp_authenticate, mcp_logout | Browser-approval auth flow for MCP clients |
| Core Memory | remember, recall, forget, update, analyze, consolidate, memory_stats | Essential memory CRUD, analysis, consolidation, and statistics |
| Discovery | search, fetch, get_guide | Tool-catalog search + fetch by ID, plus onboarding guides and personas |
| Collections | create_collection, list_collections, get_collection, add_to_collection, remove_from_collection, update_collection, delete_collection, consolidate_collection | Organize memories into groups |
| Skills & Scheduling | skill_create, skill_list, skill_render, skill_execute, skill_schedule, skill_schedule_list, skill_schedule_cancel, skill_schedule_retrigger, skill_tasks_pending, skill_task_retrigger, skill_tasks_list | Define, run, and cron-schedule pipelines |
| Tool Registry | tool_create, tool_update, tool_call, tool_list, tool_render | Register and invoke HTTP/MCP tools used by skills |
| Toolkit Search | toolkit_search | Semantic discovery across skills + tools |
| Credential Vault | vault_store, vault_list, vault_rotate | Encrypted credential storage with rotation |
| Berrry Integration | berrry_register_tool, berrry_create_tool | Register and create Berrry-app tools |
| Agents | create_agent, list_agents, agent_memory, agent_collaboration, agent_error_learning, agent_reflection, update_agent_name | Multi-agent coordination, collaboration, and learning |
Getting Started
Start with the core memory tools (remember, recall) and gradually explore advanced features like skills, collections, and the credential vault. See the full documentation for detailed usage examples.
Authentication
REST APIThe primary way to use JauMemory is through the MCP tools above. For programmatic HTTP access, MCP connectors authenticate with OAuth 2.1 + PKCE; account endpoints power the mem.jau.app web app.
https://mem.jau.app # HTTPS API + OAuth + dashboard
mem.jau.app:50051 # gRPC over TLS (used by the MCP server)
GET /.well-known/oauth-authorization-server # RFC 8414 discovery
GET /v1/authorize # Authorization endpoint (PKCE S256 required)
POST /v1/token # Token endpoint (authorization_code, refresh_token)
POST /register # Dynamic client registration (RFC 7591)
POST /v1/auth/register # Create account (requires a valid invite code)
POST /v1/auth/login # Web-app session login
POST /v1/auth/refresh # Refresh session token
GET /v1/health # Service health check
MCP tool authentication never uses passwords or API keys in chat: mcp_login returns a browser approval link, and mcp_authenticate completes the flow with the short code from the approval page.
Rate Limits & Quotas
Quotas are enforced per plan. All plans are free during the beta.
| Plan | Requests/minute | Memories/month | Collections & Agents |
|---|---|---|---|
| Hobbyist | 15 | 1,500 | 3 collections/mo · 5 agents/quarter |
| Pro | 120 | 20,000 | Unlimited |