Integrating Your First Bonds: Giving Your AI Workforce Hands

Integrating Your First Bonds: Giving Your AI Workforce Hands

Your atoms are executing tasks. Projects are moving. But right now, everything your workforce produces lives inside Supanova — documents, drafts, analysis, plans. Text deliverables.

Bonds change that. A bond is a connected integration — a tool your AI workforce can use to interact with external services. Gmail, Slack, Google Sheets, your CRM, Notion, GitHub. When you connect a bond, your atoms stop just writing about work and start doing it: sending emails, creating documents, updating records, posting messages.

This guide walks through connecting your first bond, controlling what agents can do with it, and how integrations complete the execution loop from strategy to real-world action.

What Bonds Do

Without bonds, an atom assigned to "Draft and send Q2 email campaign to enterprise segment" will draft the email and present it to you. You'd then copy it into your email tool and send it yourself.

With a Gmail bond connected, that same atom drafts the email and sends it. You review the output in the bond outputs section — subject line, recipients, body, send confirmation — and the work is done.

Bonds are the difference between an AI that writes about work and an AI that does work.

Where to Connect: The Lab

Navigate to "The Lab" from the sidebar. This is your integration hub with two views:

• Catalog: Browse all available integrations, searchable and filterable by category — productivity, communication, CRM, development, marketing, storage, analytics, and more
• Active: See and manage your currently connected integrations, check usage stats, and adjust access levels

The catalog presents a unified view. Behind the scenes, Supanova connects to 500+ external services, but you don't need to think about which provider handles what. You just find the tool you want and click connect.

Connecting Your First Bond

Most integrations use OAuth — the standard "Sign in with Google/Slack/etc." flow you've used before.

OAuth flow (most common):

1. Find the integration in the catalog (e.g., Gmail, Slack, Google Sheets)
2. Click "Connect"
3. Your browser redirects to the service's authorization page
4. Sign in and grant Supanova permission to access your account
5. Redirect back to The Lab — connection finalized automatically
6. Status changes to connected

API key flow (some integrations):

1. Click "Connect"
2. A dialog asks for your API key
3. Paste your key
4. The system validates and encrypts it
5. Connection established

That's it. Once connected, every atom in your workspace can use that integration when executing tasks — subject to the access level you set.

Access Levels: Controlling What Atoms Can Do

Every bond connection has an access level you control. This is how you set boundaries on what your workforce can do with each tool.

Read: Atoms can query and retrieve data — read emails, view spreadsheets, list CRM records — but cannot modify anything. Good for research tasks and data gathering.

Read + Create: Atoms can query data AND create new items — send emails, add spreadsheet rows, create documents, post messages — but cannot modify or delete existing items. This is the default and covers most use cases.

Full: Atoms can read, create, modify, and delete. Rarely needed. Use this only when tasks specifically require updating or removing existing records.

You can change the access level anytime from the Active tab in The Lab. Start with read + create. If you want to be cautious while you build trust with the system, start with read-only and upgrade after you've seen how atoms use the integration.

How Atoms Use Bonds During Task Execution

When an atom picks up a task, here's what happens with bonds:

1. The system checks if bonds are enabled for your workspace
2. It fetches all your active bond connections
3. It builds a context that describes what each bond can do — for example, "Gmail (Read + Create): you may query emails and send new messages"
4. That context is included in the atom's instructions
5. The atom decides which bonds to use based on what the task requires

The atom doesn't blindly use every bond available. If a task is "Write a competitive analysis report," it won't send an email just because Gmail is connected. It uses what's relevant.

All bond actions are logged with execution details and results, so you always have a record of what happened.

Bond Outputs: Seeing What Your Atoms Did

Every bond action produces an output, viewable in the project detail view under "Bond Outputs." This is your audit trail.

Output types include:

• Record: Structured data — CRM record created, spreadsheet row added
• File: Document, PDF, or image generated or uploaded
• Message: Email sent, Slack message posted
• Confirmation: Decision or approval completed
• Batch: Multiple items processed — like a bulk email send

Each output includes:

• A summary of what happened (e.g., "Sent email to john@example.com with subject 'Q2 Proposal'")
• External reference links — click to see the actual email, document, or record in the source system
• A preview of the content
• Undo capability when available

Bond outputs are how you verify that your workforce is doing what you expect. Review them after tasks complete, especially early on while you're building confidence in the system.

When an Atom Needs a Bond You Haven't Connected

If a task requires an integration you haven't set up, the task enters needs_input status with an operational gate. You'll see an orange indicator telling you:

• Which integration is needed (e.g., "Salesforce")
• What access level is required
• What the atom is trying to do

Your options:

• Connect it: Go to The Lab, set up the integration, come back and confirm. The atom retries with the new tool available.
• Skip it: Let the atom proceed without the integration. It'll use fallback approaches — producing the deliverable as text instead of executing it in the external system.

Operational gates are the system's way of telling you what tools would make the work better without forcing you to connect anything you're not ready for.

Which Bonds to Connect First

If you're not sure where to start, think through the tools you use most and connect to those.

One integration connected to start is enough. You'll see immediate value the first time an atom completes a task end-to-end — from reading the brief to delivering the output in your actual tool — without you touching anything in between.

Security and Credentials

All credentials — OAuth tokens, API keys, refresh tokens — are encrypted before storage. Supanova never stores credentials in plain text.

You can disconnect any bond instantly from the Active tab. If an OAuth token expires, the system notifies you to re-authorize. Connection statuses are visible at a glance: connected, disconnected, expired, or error.

You're always in control of what's connected and what level of access your workforce has.

How Bonds Complete the Training Loop

Bonds close the gap between planning and execution — and that matters for workforce training.

Without bonds, the platform learns from text outputs: was the draft good? Did the analysis make sense? That's useful, but limited.

With bonds, the platform learns from real outcomes: was the email opened? Did the CRM record get created correctly? Was the Slack message posted to the right channel? Real-world feedback is richer than text feedback, and it makes the learning loop tighter.

Every bond action your atoms take — successful or not — feeds back into the system's understanding of how to execute similar tasks in the future. Which email formats get responses. Which report structures your team actually uses. Which CRM fields matter for your workflow.

This is the final piece of the cascade: Objectives set direction. Projects create plans. Tasks break down work. Bonds let your workforce act on the real world. And every action teaches it to act better next time.

Tips for Your First Bond

1. Start with one integration you use daily. You'll see the value immediately when an atom delivers work directly into a tool you're already in.

2. Set access to read-only first if you're cautious. You can always upgrade to read + create after you've seen how atoms interact with the tool. There's no penalty for starting conservative.

3. Check bond outputs after tasks complete. Especially for the first few tasks that use a new integration. Verify the atom used it correctly before trusting it on higher-stakes work.

4. Monitor usage stats in the Active tab. If a bond shows zero usage, your current tasks might not need it — that's fine. Don't connect integrations speculatively.

5. Respond to operational gates promptly. When an atom requests an integration, it's telling you it could do more with better tools. Connecting the requested bond often turns a partial deliverable into a complete one.

The Full Picture

Over the course of these four guides, you've seen the complete execution model:

1. OKRs set strategic direction — your objectives tell the workforce what matters
2. Projects create execution plans — the platform breaks objectives into structured work
3. Tasks are the units of execution — atoms pick them up, execute them, and deliver results
4. Bonds connect your workforce to the real world — turning text deliverables into real actions

Each layer trains the next. Better objectives produce better projects. Better projects produce better tasks. Better tasks with better tools produce better outcomes. And better outcomes feed back into every layer, making the whole system sharper over time.

Your AI workforce is live. Set direction. Let it move.

supanova.team

... Six