Description
n8n-nodes-payi
n8n community node for Pay-i — add cost tracking, budget enforcement, and usage attribution to every LLM call in your n8n workflows.
What is Pay-i?
Pay-i is an AI cost management platform. It sits as a transparent proxy between your application and LLM providers (OpenAI, Anthropic, Azure OpenAI, AWS Bedrock, Databricks, and more), giving you:
- Real-time cost visibility — See the dollar cost of every LLM request as it happens, broken down by input/output tokens
- Budget enforcement — Set hard spending limits per user, team, use case, or workflow so costs never run away
- Usage attribution — Know exactly who and what is driving your AI spend with per-user and per-use-case tracking
- Zero code changes — Pay-i is a drop-in proxy. Your prompts, models, and provider API keys stay the same. Just route through Pay-i and you’re tracked.
Learn more at pay-i.com or read the Pay-i documentation.
Nodes
This package provides provider-specific chat model nodes and a generic proxy node:
Chat Model Nodes (LangChain-compatible)
These nodes plug directly into n8n’s AI Agent node as a chat model input. Each uses the provider’s native n8n credential type — no need to re-enter API keys.
| Node | Provider | Credential | Docs |
|——|———-|————|——|
| Pay-i OpenAI (Proxy) | OpenAI | openAiApi | docs/providers/openai.md |
| Pay-i Anthropic (Proxy) | Anthropic | anthropicApi | docs/providers/anthropic.md |
| Pay-i Azure AI Foundry (Proxy) | Azure OpenAI | azureOpenAiApi | docs/providers/azure.md |
| Pay-i Amazon Bedrock (Proxy) | AWS Bedrock | aws | docs/providers/bedrock.md |
| Pay-i Databricks (Proxy) | Databricks | payiDatabricksApi | docs/providers/databricks.md |
Generic Proxy Node
| Node | Description |
|——|————-|
| Pay-i Proxy | Direct HTTP proxy for any supported provider. Full control over the request body, provider, and model. Supports OpenAI, Anthropic, Azure OpenAI, AWS Bedrock, and Databricks. |
Installation
Community Node (Recommended)
1. In your n8n instance, go to Settings > Community Nodes
2. Select Install a community node
3. Enter n8n-nodes-payi
4. Agree to the risks and click Install
Manual Installation
cd ~/.n8n/nodes
npm install n8n-nodes-payi
Restart n8n after installing.
> AI Agent usage: To use the Pay-i Chat Model nodes with n8n’s AI Agent node, start n8n with N8NCOMMUNITYPACKAGESALLOWTOOL_USAGE=true.
Prerequisites
Setup
> Self-hosted deployment? For detailed instructions including Docker installation, dedicated Pay-i instance configuration, and troubleshooting, see the Deployment & Configuration Guide.
1. Configure Pay-i Credentials
When you first add a Pay-i node to a workflow, n8n will prompt you to configure credentials:
https://api.pay-i.com. Change this only if you’re using a self-hosted Pay-i instance.2. Configure Provider Credentials
Each chat model node uses the provider’s native n8n credential type. If you already have credentials configured for the native provider nodes (OpenAI, Anthropic, Azure, etc.), you can reuse them directly — no duplicate credential setup needed.
3. Add the Node to Your Workflow
For AI Agent workflows:
1. Add a Chat Trigger or Manual Chat Trigger node
2. Add an AI Agent node
3. Add the appropriate Pay-i [Provider] (Proxy) node and connect it to the AI Agent’s “Chat Model” input
4. Configure the model parameters (model ID, endpoint name, etc.)
Tracking Headers
Pay-i uses custom HTTP headers to associate requests with users, Use Cases, and Limits. All tracking fields are optional and have smart defaults.
| Field | Default | Description |
|——-|———|————-|
| xProxy-User-ID | (empty) | User identifier for per-user cost attribution |
| xProxy-UseCase-Name | Workflow name | Use Case Name for tracking and KPI scoring |
| xProxy-UseCase-ID | {nodeId} (UUID) | Stable per-node ID — all runs of one canvas node aggregate under one use case |
| xProxy-UseCase-Step | Node display name (e.g. Pay-i Databricks (Proxy)) | Step within a multi-step workflow — override with a custom label when useful |
Advanced Tracking (collapsed in UI)
These fields are available under the “Advanced Tracking” section and should typically only be modified with guidance from Pay-i Support:
| Field | Description |
|——-|————-|
| xProxy-UseCase-Version | Version of the use case definition |
| xProxy-UseCase-Properties | JSON object of key-value properties |
| xProxy-Limit-IDs | Comma-separated list of Pay-i Limit IDs to enforce |
| Debug Logging | Logs request URLs, headers, and routing details to the n8n server console |
Supported Proxy Paths
All requests are routed through: {PAYIBASEURL}/api/v1/proxy/{provider_path}
| Provider | Proxy Path |
|———-|———–|
| OpenAI | openai/v1/chat/completions |
| Anthropic | anthropic/v1/messages |
| Azure OpenAI | azure.openai/openai/deployments/{name}/chat/completions |
| AWS Bedrock | aws.bedrock/{region}/model/{modelId}/converse |
| Databricks | openai/v1/chat/completions (reuses OpenAI path with xProxy-Provider-BaseUri) |
Migrating Existing Workflows
Already have n8n workflows calling OpenAI, Anthropic, Databricks, or other LLM providers natively? The payi-utilities n8n-toolkit can scan your n8n instance, find native LLM nodes, and replace them with Pay-i equivalents — rewiring all connections and credentials automatically.
export N8NBASEURL=http://localhost:5678
export N8NAPIKEY=your-n8n-api-key
export PAYIBASEURL=https://api.yourcompany.pay-i.com
export PAYIAPIKEY=your-payi-api-keypython3 migrate-workflows-to-payi.py --dry-run # preview changes
python3 migrate-workflows-to-payi.py # run migration
Development
npm install # Install dependencies
npm run build # Clean, compile TypeScript, copy icons
npm run dev # TypeScript watch mode
npm run lint # ESLint (TypeScript rules)
npm run lint:fix # ESLint with auto-fix
npm run format # Prettier
Local Testing
npm pack
cd ~/.n8n/nodes && npm install /path/to/n8n-nodes-payi-*.tgz
rm -rf ~/.n8n/.cache # Clear n8n cache after updates
Restart n8n
Support
License
MIT
