n8n to JieGou Migration Guide — Step-by-Step Technical Walkthrough

This guide walks you through every step of migrating your n8n workflows to JieGou. Whether you have 5 workflows or 500, the process is the same — and most of it is automated.

Time estimate: 1–2 hours for a typical deployment of 20–50 workflows.

Prerequisites

Before you start:

• n8n access — You’ll need admin access to export workflows (self-hosted or cloud)
• JieGou account — Sign up free or deploy the Docker Compose starter kit for air-gapped environments
• Integration credentials — API keys or OAuth tokens for any services your n8n workflows connect to (Slack, Gmail, GitHub, etc.)

Step 1: Export your n8n workflows

Option A: Export all workflows (recommended)

1. Open your n8n instance
2. Go to Settings → Export All Workflows
3. Download the JSON file — it contains every workflow in your instance

The exported file looks like this:

[
  {
    "name": "Lead Enrichment Pipeline",
    "nodes": [
      { "name": "Webhook", "type": "n8n-nodes-base.webhook", ... },
      { "name": "HTTP Request", "type": "n8n-nodes-base.httpRequest", ... },
      { "name": "Set", "type": "n8n-nodes-base.set", ... }
    ],
    "connections": { ... }
  },
  ...
]

Option B: Export individual workflows

1. Open the workflow you want to migrate
2. Click the ⋮ menu → Download
3. Save the JSON file

You can import workflows one at a time or batch them.

What gets exported

Included	Not Included
Workflow name & structure	Credential values (API keys, tokens)
Node types & configuration	Execution history
Connections between nodes	Environment variables
Tags & annotations	Custom node packages
Static parameters

Important: n8n never exports credential values for security reasons. You’ll configure new credentials in JieGou as MCP server connections.

Step 2: Sign up for JieGou

Cloud (fastest)

1. Go to console.jiegou.ai
2. Sign up with Google, GitHub, or email
3. Choose your department (the import tool works from any department)
4. You’ll land on the dashboard — navigate to Integrations → Migrations

Self-hosted (Docker Compose)

For regulated environments that need data to stay on-premises:

# Clone the starter kit
git clone https://github.com/JieGouAI/self-hosted.git
cd self-hosted

# Configure
cp .env.example .env
# Edit .env with your Firebase credentials and LLM API keys

# Start
docker compose up -d

The console will be available at http://localhost:3000.

Hybrid VPC deployment (Enterprise)

For teams that need cloud convenience with on-premises data processing:

1. Contact sales@jiegou.ai for VPC agent setup
2. The control plane (console, scheduling, monitoring) runs in JieGou cloud
3. Workflow execution happens inside your VPC — data never leaves your network
4. See our hybrid deployment blog post for architecture details

Step 3: Use the import wizard

1. Navigate to /migrations/n8n in your JieGou console
2. You’ll see the n8n Import Wizard — a three-step process

Upload

• Paste your n8n workflow JSON directly into the text area, or
• Upload the .json file using the file picker

Click Preview Conversion to analyze the workflow.

What happens during analysis

The import tool:

1. Parses the n8n workflow JSON (validates structure, extracts nodes and connections)
2. Maps each n8n node type to the closest JieGou step type
3. Detects trigger nodes (webhooks, schedules, manual triggers) and converts them to workflow inputs
4. Identifies integration nodes and suggests matching MCP servers
5. Builds the connection graph (preserving execution order and branching logic)
6. Generates a conversion report with warnings for anything that needs manual attention

Step 4: Review and adjust mappings

After analysis, you’ll see a detailed conversion report.

Summary cards

The top of the report shows four metrics:

• Total Nodes — How many n8n nodes were in the workflow
• Mapped — Nodes that converted cleanly to JieGou step types
• Skipped — Nodes that don’t need conversion (e.g., NoOp, StickyNote)
• Warnings — Items that need your attention

Node mapping table

Each n8n node is shown with its conversion status:

Status	Meaning	Action Required
Mapped	Cleanly converted to a JieGou step	None — verify the step looks right
Partial	Converted but some parameters need manual review	Check the step configuration
Manual	No automatic conversion available	Build the equivalent step manually
Skipped	Node type doesn’t need conversion	None

How n8n nodes map to JieGou steps

n8n Node	JieGou Step	Notes
Set / Code / Function	LLM Step	AI performs the transformation instead of code
IF / Switch / Filter	Condition Step	Direct mapping — same branching logic
SplitInBatches	Loop Step	Direct mapping — iterates over array input
Merge / Aggregate	Aggregator Step	Combines outputs from parallel branches
HTTP Request	LLM Step + MCP Tool	Uses the HTTP MCP server for API calls
Slack / Gmail / GitHub	LLM Step + MCP Server	Uses the matching MCP server integration
Webhook (trigger)	Workflow Input	Becomes the workflow’s input schema
Schedule Trigger	Schedule	Configure separately in JieGou Schedules
OpenAI / LangChain Agent	LLM Step	Multi-provider BYOK — bring your own key
Postgres / MySQL	LLM Step + MCP Server	Uses the database MCP server
Google Sheets	LLM Step + MCP Server	Uses the Google Sheets MCP server

Understanding warnings

Warnings fall into three categories:

• 🔴 Error — Something can’t be automatically converted. Example: “Custom function node uses npm packages not available in JieGou”
• 🟡 Warning — Conversion succeeded but needs verification. Example: “Slack node detected — configure Slack MCP server for this integration”
• 🔵 Info — Helpful context. Example: “Webhook trigger converted to workflow input schema”

Common warnings and how to resolve them:

Warning	Resolution
”Configure [X] MCP server”	Go to Library → Integrations, install the MCP server, and connect your credentials
”Complex expression may need adjustment”	Review the step’s prompt/configuration — n8n expressions like {{$json.field}} are converted to input mappings
”Custom code node”	Review the generated LLM prompt — it describes what the code did, and the LLM will perform the same task
”Unsupported node type”	Build the equivalent manually or check if an MCP server provides the functionality

Step 5: Configure your MCP servers

MCP (Model Context Protocol) servers replace n8n’s built-in integration nodes. Each MCP server provides tools that LLM steps can use.

Install MCP servers

1. Go to Library → Integrations
2. Search for the integration you need (e.g., “Slack”, “GitHub”, “PostgreSQL”)
3. Click Install to add it to your account
4. Connect with your API key or OAuth credentials

Map n8n credentials to MCP servers

n8n Credential	JieGou MCP Server	Connection Method
Slack OAuth	Slack MCP Server	OAuth 2.0 (via Klavis)
Gmail API	Gmail MCP Server	OAuth 2.0 (via Klavis)
GitHub Token	GitHub MCP Server	Personal Access Token
Postgres Connection	PostgreSQL MCP Server	Connection string
HTTP Header Auth	HTTP MCP Server	API key / Bearer token
OpenAI API Key	Account BYOK Settings	API key in Key Vault
AWS Credentials	AWS MCP Server	Access key + secret
Google Sheets OAuth	Google Sheets MCP Server	Service account or OAuth

Verify tool access

After installing MCP servers:

1. Open the imported workflow
2. Check each LLM step that uses MCP tools
3. Verify the correct MCP server is selected
4. Test the connection with a quick manual run

Step 6: Test with a bakeoff

Before going live, run both your n8n workflow and JieGou workflow side-by-side to compare output quality.

Create a bakeoff

1. Go to Bakeoffs → New Bakeoff
2. Select Workflow vs. Workflow mode
3. Add your JieGou workflow as Arm A
4. For Arm B, you have two options:
   • Run your n8n workflow separately and paste the output
   • Create a second JieGou workflow variant with different LLM settings

Generate test inputs

1. Click Generate Inputs — JieGou creates synthetic test data matching your workflow’s input schema
2. Or provide real inputs from your n8n execution history
3. We recommend 5–10 diverse inputs for a meaningful comparison

Review results

The bakeoff runs both arms on every input and provides:

• LLM-as-judge evaluation — An AI evaluator scores each output on accuracy, completeness, and relevance
• Statistical confidence — Standard deviation and 95% confidence intervals
• Side-by-side comparison — View outputs next to each other for manual review

What to look for

• Output format — Does the JieGou workflow produce output in the same format your downstream systems expect?
• Data completeness — Are all fields populated? Any missing information?
• Edge cases — How do both handle empty inputs, large payloads, or unusual data?
• Error handling — What happens when an API call fails?

Step 7: Go live

Once you’re confident in the conversion:

Cut over gradually

1. Start with low-risk workflows — internal notifications, report generation, data sync
2. Monitor for 48 hours — check execution logs in JieGou’s run history
3. Move critical workflows — customer-facing integrations, payment processing, data pipelines
4. Decommission n8n — once all workflows are verified, shut down your n8n instance

Set up scheduling

If your n8n workflows used schedule triggers:

1. Go to Automate → Schedules
2. Create a new schedule for each workflow
3. Configure the cron expression (JieGou supports the same cron syntax as n8n)
4. Set the timezone and input parameters

Set up webhook triggers

If your n8n workflows used webhook triggers:

1. Go to Automate → Triggers
2. Create a new webhook trigger
3. Copy the new webhook URL
4. Update the webhook URL in your upstream systems (e.g., Stripe, GitHub, Slack)

Enable monitoring

• Execution alerts — Set up notifications for failed runs
• Audit logging — All workflow executions are automatically logged with 30 action types
• SOC 2 evidence — Export compliance evidence from the Compliance Dashboard

Troubleshooting

”Invalid JSON” error during upload

• Verify the file is valid JSON (try pasting it into jsonlint.com)
• If you exported multiple workflows, the file is a JSON array — the import tool handles both single workflows and arrays
• Check for BOM characters or encoding issues (save as UTF-8)

Node marked as “Manual Required”

This means the import tool couldn’t automatically convert this node type. Common cases:

• Custom Function nodes with complex npm dependencies → Rewrite the logic as an LLM prompt or use the HTTP MCP server to call an external API
• FTP / SFTP nodes → Use the filesystem MCP server or an HTTP-based file transfer service
• XML nodes → LLM steps handle XML parsing natively — describe the transformation in the prompt

Workflow execution fails after import

1. Check the run detail page for the specific error
2. Common causes:
   • MCP server not configured (install from Library → Integrations)
   • Input schema mismatch (check the workflow’s input configuration)
   • Missing BYOK key for the LLM provider (add in Account Settings → API Keys)

Performance differences

JieGou workflows are AI-native — instead of executing code in a VM, they use LLM steps to perform transformations. This means:

• Latency may be higher for simple data transformations (LLM calls vs. in-memory code)
• Quality may be higher for unstructured data processing (LLM understanding vs. regex/code)
• Maintenance is lower — no code to debug when APIs change

If latency is critical, use the Condition Step to skip LLM calls for deterministic operations and reserve LLM steps for tasks that benefit from AI.

Migration checklist

Use this checklist to track your migration progress:

• Export all n8n workflows as JSON
• Create JieGou account (cloud, self-hosted, or hybrid)
• Upload workflows to the import wizard
• Review conversion reports for each workflow
• Install required MCP servers from Library → Integrations
• Configure MCP server credentials
• Run bakeoffs for critical workflows
• Set up schedules for cron-triggered workflows
• Update webhook URLs in upstream systems
• Monitor execution for 48 hours
• Decommission n8n instance
• Export SOC 2 evidence for audit trail

Need help?

• Import tool: console.jiegou.ai/migrations/n8n
• MCP Integrations: console.jiegou.ai/integrations
• Community: community.jiegou.ai
• Enterprise support: sales@jiegou.ai
• Security details: Why we built hybrid deployment