Skip to main content
This is one of the most powerful things about your Spark agent: it’s not limited to a fixed list of integrations. If a tool has an API, your agent can use it. Period. Unlike traditional integration platforms where you’re stuck waiting for someone to build a connector, your agent is a general-purpose operator. Give it credentials and documentation, and it figures out how to work with the tool.

​
How It Works

Your agent can make HTTP requests, authenticate with API keys or tokens, read JSON responses, and chain multiple API calls together. It doesn’t need a pre-built connector — it needs the same things a developer would need:
  1. Credentials — an API key, access token, or OAuth connection
  2. Documentation — how the API works (endpoints, parameters, response formats)
That’s it. Your agent reads the docs and operates the tool.

​
Setting Up a Custom Integration

1

Get API Credentials

Log into the service you want to connect. Find their Developer or API section and generate an API key or access token. Most services make this straightforward.
2

Upload API Documentation

Go to Knowledge in your workspace and create a new document. Paste the relevant API documentation — you don’t need the entire thing, just the endpoints your agent will use.For example, if you want your agent to create invoices in QuickBooks, upload the docs for the Invoices API endpoints, not the entire QuickBooks developer guide.
3

Give Your Agent the Credentials

Tell your agent in chat:
“Here’s the API key for our QuickBooks account: [key]. Use the QuickBooks API documentation I uploaded to interact with our accounting data.”
Your agent stores this securely and uses it for all future requests to that service.
4

Start Using It

Just ask your agent to do things with the tool:
“Pull all unpaid invoices from QuickBooks that are over 30 days old”
“Create a new invoice in QuickBooks for Acme Corp, $5,000, due in 30 days”
Your agent reads the API docs from your knowledge file, constructs the right API call, authenticates with your credentials, and returns the results.

​
Examples of Custom Integrations

Your agent can work with virtually anything. Here are some real examples:
Create and send invoices, pull financial reports, track payments, reconcile accounts, generate P&L statements. Upload the accounting API docs and your agent becomes your bookkeeper.
Sync contacts, update deal stages, pull pipeline reports, log activities. Your agent can work alongside your existing CRM or replace manual data entry entirely.
Create tasks, update statuses, assign team members, pull project reports, track milestones. Your agent can manage project workflows across whatever PM tool you use.
Monitor orders, update inventory, track shipments, pull sales analytics, manage product listings. If the platform has an API, your agent can operate it.
Look up employee info, track PTO, pull headcount reports, monitor open positions. Your agent respects the same permission boundaries as the API key you provide.
Pull usage metrics, identify trends, generate reports, set up alerts on metric changes. Your agent can synthesize data from multiple analytics tools into unified reports.
Monitor conversations, draft responses, categorize tickets, pull customer interaction history. Connect any communication tool your team uses.
If your company has custom internal tools, databases, or microservices with API endpoints, your agent can interact with them too. Upload your internal API documentation as a knowledge file and provide the auth credentials.

​
Webhooks for Inbound Data

For services that need to send data to your agent (form submissions, payment notifications, status updates, event triggers), use webhooks. Your agent receives the data in real-time and processes it according to your instructions. This works with any service that supports webhook notifications — no built-in integration needed.

​
Tips for Custom Integrations

Keep API docs focused. Don’t upload a 200-page developer guide. Extract the 5-10 endpoints your agent actually needs. A focused knowledge file helps your agent find the right information faster.
Use least-privilege credentials. Generate API keys with only the permissions your agent needs. Read-only keys for reporting, read-write for operations. This limits blast radius if anything goes wrong.
Test with a simple request first. Before setting up complex workflows, ask your agent to make one simple API call: “List the 5 most recent invoices from QuickBooks.” This confirms credentials work and the agent can interpret the API correctly.
API rate limits apply. Most services limit how many API calls you can make per minute/hour. Your agent respects these limits, but if you set up aggressive schedules (checking every minute), you may hit them. Start with reasonable intervals and adjust.