Hindsight ships with native integration packages for the most popular agent frameworks. Each package wraps the Hindsight Python client and exposes memory as tools, nodes, or memory interfaces that fit naturally into each framework’s patterns. If your framework isn’t listed, you can connect via the REST API or the Python and Node.js SDKs directly.Documentation Index
Fetch the complete documentation index at: https://mintlify.com/vectorize-io/hindsight/llms.txt
Use this file to discover all available pages before exploring further.
MCP server
Connect Claude Code, Claude Desktop, Cursor, or any MCP-compatible tool to a memory bank without writing code
LangChain / LlamaIndex
Memory tools and native memory interfaces for LangChain, LangGraph, and LlamaIndex agents
OpenAI Agents SDK
FunctionTool instances for retain, recall, and reflect — async-native and drop-in compatibleCrewAI / AutoGen / AG2
Storage backends, memory tools, and GroupChat support for multi-agent crew frameworks
LiteLLM / n8n / Dify
Connect no-code and proxy-layer tools via the REST API or built-in MCP server
REST API & SDKs
Any framework can call retain, recall, and reflect directly using the Python SDK, Node.js SDK, or HTTP
How integrations work
Every Hindsight integration wraps the same three operations:- Retain — store information and trigger fact extraction, entity resolution, and embedding
- Recall — search memories using semantic similarity, BM25, graph traversal, and temporal reasoning
- Reflect — synthesize a grounded, disposition-aware answer from retrieved memories
Choosing an integration pattern
| Pattern | Best for |
|---|---|
| MCP server | AI assistants and IDEs (Claude, Cursor) — no code required |
| Native package | Agent frameworks (LangChain, CrewAI, OpenAI Agents) — idiomatic integration |
| REST API / SDK | Custom frameworks or fine-grained control |
All integration packages require a running Hindsight API server. See the installation guide to get started with Docker, Helm, or the embedded mode.