Integrations
Dakera publishes integration packages for LangChain (Python and JavaScript), CrewAI, LlamaIndex and AutoGen. Each wraps the Dakera REST API in components that fit the framework: memory backends, vector stores, session managers and knowledge-graph helpers. This page helps you pick a package; each package has its own page with installation and examples.
All five packages are at 0.3.0 and call the server through the Dakera SDKs (Python or JavaScript), so the v0.12.0 notes on the Python SDK and TypeScript SDK pages apply to them too.
Supported frameworks
| Framework | Package | Language | Memory | VectorStore | Sessions | Graph |
|---|---|---|---|---|---|---|
| LangChain | langchain- |
Python | ✓ | ✓ | ✓ | ✓ |
| LangChain.js | @dakera- |
TypeScript | ✓ | ✓ | ✓ | ✓ |
| CrewAI | crewai-dakera |
Python | ✓ | ✓ | ✓ | |
| LlamaIndex | llamaindex- |
Python | ✓ | ✓ | ✓ | ✓ |
| AutoGen | autogen-dakera |
Python | ✓ | ✓ | ✓ |
Getting started
All integrations follow the same pattern:
- Run Dakera — start the server with Docker or from source
- Install the package —
pip install <package>ornpm install <package> - Configure — pass the server URL and API key to the component (the packages do not read environment variables themselves)
- Use — import the framework-native class and pass it to your chain/agent
Choosing an integration
Using Claude Desktop or Cursor? Use the Cursor MCP guide or Claude Desktop MCP guide — connect Dakera's 14 core memory tools directly to your IDE without writing any code.
Building with LangChain? Use LangChain or LangChain.js — each provides conversation memory and a VectorStore implementation.
Multi-agent orchestration? Use CrewAI or AutoGen — each agent gets its own persistent memory, keyed by agent_id.
RAG pipeline? Use LlamaIndex — DakeraIndexStore is a LlamaIndex vector store with server-side embeddings.
LLM governance (cost tracking, policy decisions)? Use TealTiger — DakeraCostStorage and DakeraDecisionStore for decay-weighted governance state.
Common configuration
Every component takes the server URL and an API key. The examples on the integration pages read them from these environment variables (the packages themselves do not):
| Variable | Description | Default |
|---|---|---|
DAKERA_API_URL | Dakera server URL | http:// |
DAKERA_API_KEY | API key for authentication | (none) |
All integrations
- Cursor (MCP) — 14 core memory tools for Cursor IDE via MCP protocol
- Claude Desktop (MCP) — persistent memory for Claude Desktop via MCP
- LangChain (Python) — DakeraMemory + DakeraVectorStore
- LangChain.js (TypeScript) — DakeraMemory + DakeraVectorStore
- CrewAI — DakeraStorage for persistent crew memory
- LlamaIndex — DakeraMemoryStore + DakeraIndexStore
- AutoGen — DakeraMemory for multi-agent conversations
- TealTiger (Governance) — DakeraCostStorage and DakeraDecisionStore
Pick your stack and start
Every integration runs against the same self-hosted Dakera server. Start with the Quick Start to get the server running, then add the integration that fits your stack.