Framework Adapters
Drop Rekall into your agent framework of choice. Each adapter implements the framework's own memory, session, or checkpoint contract and routes it to Rekall's hosted memory and execution-state layers — so durable, cross-host memory is shared across all your agents and editors.
Overview
Framework adapters are thin clients of Rekall's existing REST endpoints — /api/v2/memories for memory and /api/v1/execution-memory/external for durable state — so they require no backend changes. They are duck-typed against their target framework (no hard version coupling) and authenticate with the same rk_ API key or rat_ access token you use everywhere else.
One store, every agent
Because every adapter writes to the same Rekall account, memory captured by your LangGraph graph is recallable from your CrewAI crew, your Claude agent, and your editor plugins. Pick the adapters you need and they share context automatically.
TypeScript Adapters
LangGraph
A LangGraph checkpointer backed by Rekall execution memory. Give your graphs durable, server-persisted state — resume after a crash, time-travel, and human-in-the-loop — in one line.
@rekall/langgraphLearn more Mastra
Rekall memory plus durable session state for Mastra, delivered as an input/output Processor: recall-injection into the prompt and capture of the assistant output back into memory.
@rekall/mastraLearn more OpenAI Agents
A durable Rekall-backed Session for the OpenAI Agents SDK — conversation memory that survives restarts and moves across hosts, in one object.
@rekall/openai-agentsLearn more Vercel AI SDK
Rekall memory plus durable session state for the Vercel AI SDK. Recall context into your prompts, capture what the model produces, and persist conversation state across runs.
@rekall/vercel-aiLearn more Claude Agent SDK
Rekall memory plus durable session state for the Claude Agent SDK. Recall relevant context, capture what agents learn, and keep state that survives restarts.
@rekall/claude-agentLearn more Python Adapters
CrewAI
Rekall-backed memory for CrewAI 1.x via the StorageBackend protocol. Durable, cross-session crew memory shared across runs, hosts, and agents — with MCP recommended for best-in-class recall.
rekall-crewaiLearn more AutoGen
Rekall-backed memory for AutoGen — implements the autogen_core.memory.Memory protocol so an AgentChat agent retrieves relevant facts from Rekall and persists what it learns.
rekall-autogenLearn more Pydantic AI
Rekall memory plus durable message history for Pydantic AI. Persist a conversation across runs and add semantic recall on top.
rekall-pydantic-aiLearn more Choosing an Adapter
Adapters fall into two broad shapes. Some persist durable execution state (a graph checkpoint, a session's message history) so an agent can resume across crashes and hosts. Others add semantic memory — recall relevant facts into the prompt and capture what the model learns. Several do both.
| Adapter | Durable State | Semantic Memory | Framework Hook |
|---|---|---|---|
| LangGraph | Yes | — | Checkpointer (Saver) |
| OpenAI Agents | Yes | — | Session |
| CrewAI | Yes | Cosine (StorageBackend) | StorageBackend |
| AutoGen | — | Yes | Memory protocol |
| Mastra | Yes | Yes | Processor |
| Pydantic AI | Yes | Yes | Message history + recall |
| Vercel AI SDK | Yes | Yes | System prompt + callbacks |
| Claude Agent SDK | Yes | Yes | System prompt + hooks |
