# MemexAI

- [MemexAI Docs](/docs): Memory that changes your agent's next response, backed by Postgres.
- [Architecture](/docs/architecture): How MemexAI is deployed, how tools flow, and where memory lives.
- Quickstart
  - [Coding Agent Onboarding](/docs/quickstart/agent-onboarding): Let a coding agent set up MemexAI from a public setup file.
  - [Containerized Service](/docs/quickstart/docker-service): Run MemexAI as a separate service, then connect over TypeScript, Python, or MCP.
  - [Advanced: Direct Postgres Runtime](/docs/quickstart/direct-postgres): Skip the service only when your app intentionally owns database credentials.
- [MCP Clients](/docs/mcp): Connect Model Context Protocol clients over SSE or stdio.
- SDKs
  - [TypeScript SDKs](/docs/sdks/typescript): Use @memexai/sdk for the service or @memexai/core as an in-process runtime.
  - [Python SDK](/docs/sdks/python): Connect Python apps to the recommended MemexAI service.
- Adapters
  - [Vercel AI SDK](/docs/adapters/vercel-ai): Use MemexAI memory tools with the Vercel AI SDK in TypeScript.
  - [Anthropic SDK](/docs/adapters/anthropic): Use MemexAI memory tools with the Anthropic SDK in TypeScript.
  - [LangChain](/docs/adapters/langchain): Use MemexAI memory tools with LangChain in TypeScript and Python.
  - [OpenAI SDK](/docs/adapters/openai): Use MemexAI memory tools with the OpenAI SDK in TypeScript.
  - [LlamaIndex](/docs/adapters/llamaindex): Use MemexAI memory tools with LlamaIndex in Python.
  - [CrewAI](/docs/adapters/crewai): Use MemexAI memory tools with CrewAI in Python.
- Concepts
  - [Design Principles](/docs/concepts/design-principles): The decisions that shape how MemexAI stores, retrieves, and isolates memory — and why.
  - [How MemexAI Works](/docs/concepts/how-it-works): The mental model for memory files, tool calls, prompt injection, and later recall.
  - [Memory Tools](/docs/concepts/memory-tools): Choose between agentic memory tools and raw file-level tools.
  - [Prompt Block](/docs/concepts/prompt-block): Make stored memory part of the system prompt so the next response can change.
  - [Shared Memory as Behavior Guide](/docs/concepts/shared-memory): Use shared memory files as a knowledge base, operations manual, tool policies, and durable behavioral guidance that every agent reads automatically.
  - [Memory Scopes](/docs/concepts/scopes): How virtual paths isolate user memory and shared memory.
  - [Revision History](/docs/concepts/revisions): Every write creates a full snapshot for auditability.
  - [Access Logs](/docs/concepts/access-logs): Track reads and writes performed by agents.
- Operations
  - [Admin Console](/docs/operations/admin-console): Inspect, correct, and operate MemexAI memory files, revisions, access logs, observability, tools, and dreaming.
  - [Background Dreaming](/docs/operations/dreaming): Configure and operate background memory consolidation for long-horizon agent memory.
  - [Production Considerations](/docs/operations/production-caveats): Honest early-stage limits to review before deploying MemexAI in production.
  - [User ID Trust Model](/docs/operations/trust-model): How MemexAI treats userId, API keys, admin secrets, and memory isolation boundaries.
  - [Correction Workflow](/docs/operations/correction-workflow): How to inspect, edit, and verify a wrong memory in MemexAI.
  - [Migrations](/docs/operations/migrations): How MemexAI manages its Postgres schema.