Memory for the AI Agent in n8n.
n8n-nodes-niadra is an n8n community node. Put it before and after your AI Agent node, or give it to the agent as a tool, so the agent starts every conversation knowing what the customer already told your other agents, and so what it says reaches them too. No code.
The n8n AI Agent's memory is this execution's message window, or a per-session store of your choice. It belongs to the workflow. Another vendor's voice agent, another workflow's WhatsApp agent and the agent acting in the ERP do not read what this workflow knows, and it does not read what they did.
The node reads the customer's context before the agent, by the identifier your trigger brings (the WhatsApp wa_id, the caller's number, your user id), and records what the customer and the agent said after. Every operation fails open by default: with Niadra slow or down, the node returns an empty result with an error field and the workflow goes on.
Install
In n8n, Settings, Community Nodes, Install, and type n8n-nodes-niadra. The package lives in packages/n8n-nodes-niadra of the TypeScript SDK repository and is published to npm next to @niadra/sdk. The Niadra API credential is a source key of your space (nia_sk_...): it names the region and the space, so the address comes from it; Base URL only for the local emulator.
How the adapter wires in
The five primitives of every Niadra integration, in this framework's extension points.
- Context
- Get Context returns the customer's context for the agent's prompt: text, for after your system message, and suffix, with the live turns from other channels, for the end of the prompt.
- Turns
- Track Turn records what the customer or the agent said, keyed by the provider's message id, so a retried step is recorded once. End records that the conversation ended.
- Tools
- Search History searches past conversations, promises, actions and business objects; When takes the period in the customer's words ("last week", "semana passada"). The node can also be given to the AI Agent as a tool.
- Verification
- Verify records that the customer proved who they are: an OTP, a login, the carrier's attestation.
- Handoff
- Handoff records a transfer to a person or to another agent.
What the agent receives
The context is compiled when the memory changes and served ready, with no AI model on the read. What another channel said during the conversation arrives as a delta, at the end of the prompt.
- Who the customer is, by what the conversation has proven: the verification level decides what goes in
- Facts, open items and promises, with the date and the channel they came from
- What other agents did inside the company, confirmed by the system of record
- Patterns computed by rule, with the evidence and the expiry
- The three history tools: search, timeline and open an item, bound to the customer in your code
- A receipt of every read, chained by SHA-256
What the adapter does not do
- Every operation fails open by default: when Niadra is slow or down, the node returns an empty result with an error field and the workflow goes on. Turn Fail Open off to stop the workflow instead.
- Bind Customer Handle to your trigger's data (the WhatsApp wa_id, the caller's number, your user id), never to a value the model writes: the customer is never something the model chooses.
- The node carries neither agent memory nor the timeline; for those, use the SDK or the MCP server.
- Built and tested against n8n-workflow 2.16. The layout of the node and its credential follows Mem0's n8n node (MIT), credited in NOTICE.
Frequently asked questions
How do I install it?
In n8n, Settings, Community Nodes, Install, and type n8n-nodes-niadra. The credential is a source key of your space (nia_sk_...): it names the region and the space, so the address comes from it. Base URL only for the local emulator.
Where does the node go in the workflow?
Get Context before the AI Agent, with the Customer Handle coming from the trigger; Track Turn after, with the customer's message and the agent's answer; End when the conversation ends. Search History can go as a tool of the agent.
What if Niadra does not answer in the middle of the workflow?
The node returns an empty result with an error field and the workflow goes on: the agent answers without the context, as it does today. If you would rather the workflow stopped, turn Fail Open off on the node.
Do I have to change my model, my prompt or my vendor?
No. The adapter places the context after your instructions and the delta at the end of the prompt, in the extension points the framework already has. Your model, your prompt and your vendor stay the same, and switching any of them later does not erase the memory.
Where does the data live, and what does it cost?
The data stays in a single region, stated in the contract, encrypted with AES-256-GCM under a key exclusive to your company and protected in a FIPS 140-3 HSM. The price is per conversation or task in which an agent read the memory: US$ 2 to 3 per thousand, by volume, with reads, searches and system events included. Niadra is opening to companies by request, before the public launch.
Tell us what you are building.
A work email and two lines about your agents are enough. The people who write the code reply, with an early-access proposal for your case.
Rather tell us more about your company? Use the full form