# Rembr - Memory Infrastructure for AI Agents

## Service Overview

Rembr is a production-ready memory service for AI agents built on the Model Context Protocol (MCP). It solves the "Context is Consciousness" problem by providing persistent, semantic memory across sessions, identity continuity during model switches, and temporal debugging capabilities.

**Tagline:** Memory Infrastructure for AI Agents
**Version:** 1.4.0
**Protocol:** Model Context Protocol (MCP)
**Documentation:** https://rembr.ai/docs

## MCP Endpoint

- **Endpoint:** `https://rembr.ai/mcp`
- **Protocol:** MCP (Model Context Protocol)
- **Authentication:** OAuth 2.0, API Key, or JWT
- **Tools:** Discover the current MCP tool suite with `tools/list`
- **Content-Type:** `application/json`

## Authentication Summary

Rembr supports multiple authentication methods:

1. **OAuth 2.0 (Recommended)** — Dynamic client registration via RFC 7591, authorization code flow with PKCE (S256)
2. **API Key** — Long-lived key prefixed `mb_live_` for server-to-server auth
3. **JWT** — Short-lived, purpose-bound internal delegation tokens

**OAuth Metadata:** `https://rembr.ai/.well-known/oauth-authorization-server`

## Core Capabilities

### Memory Intelligence
- **Hybrid Search:** Semantic + keyword + phrase matching with bounded, scoped results
- **Contradiction Detection:** Flags conflicting memories automatically
- **Relationship Inference:** Automatic semantic connections between memories
- **Temporal Debugging:** Point-in-time state queries
- **Runtime Discovery:** Authenticated `tools/list` is authoritative for the credential and release

### Architecture
- Tenant, project and user authorisation, with FORCE RLS on memories and selected MCP tables
- TLS in transit; at-rest encryption depends on verified storage configuration
- Scoped audit records for supported security-sensitive operations
- MCP-compliant protocol (not proprietary)

## Tool Categories

- **Memory Operations:** Create, read, update, delete memories
- **Search & Discovery:** Semantic search, phrase matching, relationship queries
- **Graph Operations:** Memory graph visualization, relationship management
- **Snapshot Management:** Create, retrieve, and list bounded authorised snapshots
- **Statistics:** Credential-scoped usage and embedding statistics where enabled

## Agent Integration

- **Agent Signup:** `POST https://rembr.ai/api/agent-signup` with `{"agent_name": "your-name"}`
- **Returns:** apiKey, tenantId, claimUrl
- **Free Tier:** 1,000 memories, no credit card required
- **Agent Guide:** https://rembr.ai/docs/agent-guide
- **Manifest:** https://rembr.ai/agents.json

## Security Model

- MCP-compliant protocol with proper access controls
- Tenant, project and user authorisation, with selected FORCE RLS boundaries
- TLS in transit; at-rest encryption depends on verified storage configuration
- Scoped audit records visible only to authorised operators
- Public guidance does not ask agents to override operators or conceal actions
- All agent-facing content in public source code

## Sitemap

https://rembr.ai/sitemap.xml