> ## Documentation Index
> Fetch the complete documentation index at: https://manifest.build/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# n8n

> Use Manifest as the chat model behind n8n AI Agents, or call it directly from a workflow, with the n8n-nodes-manifest community node.

<div className="deploy-mode-toggle" role="group" aria-label="Deployment mode">
  <button type="button" data-deploy-mode="cloud">Cloud</button>
  <button type="button" data-deploy-mode="selfhosted">Self-hosted</button>
</div>

Manifest ships as a verified n8n community node, [`n8n-nodes-manifest`](https://www.npmjs.com/package/n8n-nodes-manifest). It installs from n8n's own **Community Nodes** screen on n8n Cloud and on self-hosted n8n alike, and gives you two nodes:

<CardGroup cols={2}>
  <Card title="Manifest Chat Model" icon="bot">
    A language-model sub-node. Plug it into the **AI Agent** or **Basic LLM Chain** node in place of the OpenAI Chat Model, and Manifest routes every request the agent makes.
  </Card>

  <Card title="Manifest" icon="plug">
    An action node for calling Manifest directly from a workflow: list models, create a chat completion, or create a Responses API call.
  </Card>
</CardGroup>

Both nodes are plain HTTP clients for the [gateway](/docs/llm-gateway). Nothing about routing, limits, or [Autofix](/docs/autofix) changes because the caller is n8n.

## Install the node

<Steps>
  <Step title="Open Community Nodes">
    In n8n, go to **Settings** > **Community Nodes** and select **Install**.
  </Step>

  <Step title="Enter the package name">
    Enter `n8n-nodes-manifest`, confirm the community-node notice, and install.
  </Step>

  <Step title="Find the nodes">
    Search the node panel for **Manifest**. The Chat Model sub-node appears where n8n lists language models, and the action node appears with the regular nodes.
  </Step>
</Steps>

The Chat Model sub-node needs an n8n version that ships `@n8n/ai-node-sdk` (2026 releases). The action node works on older versions.

## Create the credential

Both nodes share one **Manifest API** credential with two fields.

<div data-deploy="cloud">
  | Field        | Value                                                                       |
  | ------------ | --------------------------------------------------------------------------- |
  | **Base URL** | `https://app.manifest.build`                                                |
  | **API Key**  | The harness key, starting with `mnfst_`, from the harness **Settings** page |
</div>

<div data-deploy="selfhosted">
  | Field        | Value                                                                       |
  | ------------ | --------------------------------------------------------------------------- |
  | **Base URL** | The origin of your instance, for example `http://192.168.1.20:2099`         |
  | **API Key**  | The harness key, starting with `mnfst_`, from the harness **Settings** page |
</div>

The Base URL is the origin only, without `/v1`. The nodes append `/v1/models`, `/v1/chat/completions`, and `/v1/responses` themselves. A Base URL ending in `/v1` sends requests to `/v1/v1/...` and every call fails.

Select **Test** on the credential. It calls `GET /v1/models`, so a passing test confirms that n8n can reach your instance and that the key is valid.

<div data-deploy="selfhosted">
  ### Reaching a self-hosted instance from n8n

  The credential test fails with a connection error when n8n cannot reach Manifest. Where the two run decides the address:

  * **n8n and Manifest in Docker on the same host.** `localhost` inside the n8n container is n8n itself. Put both containers on a shared Docker network and use the Manifest service name as the host, or use `http://host.docker.internal:2099` on Docker Desktop.
  * **n8n on another machine.** The bundled compose file binds Manifest to `127.0.0.1` only. Set `HOST_BIND_ADDRESS=0.0.0.0` in Manifest's `.env`, as described in [Exposing on the LAN](/docs/self-hosted#exposing-on-the-lan), so the port answers on the network.
  * **Plain `http://` is fine.** The calls are server to server. The dashboard's browser security rules do not apply to them.

  The keyless loopback shortcut that a development build of Manifest accepts does not exist in the production image. n8n always needs a real `mnfst_` key.
</div>

## Use Manifest behind an AI Agent

<Steps>
  <Step title="Add the Chat Model">
    On an **AI Agent** or **Basic LLM Chain** node, select the **Model** connector and choose **Manifest Chat Model**.
  </Step>

  <Step title="Pick the credential and model">
    Select the Manifest credential. The model list is loaded from your instance. Leave it on `auto` to let your [routing rules](/docs/llm-gateway#routing) pick the model, or choose a model ID to pin one.
  </Step>

  <Step title="Run the workflow">
    Streaming and tool calling work as they do with the OpenAI Chat Model, so agent tools run unchanged. Each call appears on the harness's **Requests** page with the model that served it and its cost.
  </Step>
</Steps>

An option on the sub-node switches it from Chat Completions to the [Responses API](/docs/reference/api#endpoints). Leave it off unless a model you route to needs Responses.

## Call Manifest from a workflow

The **Manifest** action node has three operations:

| Operation                  | Gateway route               |
| -------------------------- | --------------------------- |
| **List Models**            | `GET /v1/models`            |
| **Create Chat Completion** | `POST /v1/chat/completions` |
| **Create Response**        | `POST /v1/responses`        |

For a first test, choose **Create Chat Completion**, set **Model** to `auto`, and use these messages:

```json theme={"theme":{"light":"github-light","dark":"github-dark"}}
[
  { "role": "user", "content": "Reply with one sentence confirming that Manifest is connected." }
]
```

Your routing decides whether the route answers with buffered JSON or a stream. The node handles both: a buffered response is returned as the API's JSON object, and a streamed response is collected until it finishes and returned as its parsed server-sent events. Do not set `stream` in **Additional Body**. The node ignores that field.

## Troubleshooting

| Symptom                                                                                           | Cause                                                                                                            |
| ------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------- |
| Credential test fails with a connection error                                                     | n8n cannot reach the Base URL. See [Reaching a self-hosted instance](#reaching-a-self-hosted-instance-from-n8n). |
| Credential test fails with `401`                                                                  | The key is not a valid `mnfst_` harness key. Copy it again from the harness **Settings** page.                   |
| Every call returns `404`                                                                          | The Base URL ends in `/v1`. Remove it.                                                                           |
| The model list shows only `auto`, or calls fail with [M100](/docs/errors/M100) or [M101](/docs/errors/M101) | No provider is connected to the harness yet. Connect one on the harness's **Providers** tab.                     |