Connecting to a Remote Pega Application with Agentic Process Fabric (APF)
Agentic Process Fabric lets one Pega instance discover and orchestrate agents running on a completely separate, remote Pega application — handing off work like remote case creation without the two systems needing to share a codebase. Here’s the end-to-end flow for setting this up, plus a gotcha that can trip you up even after everything looks configured correctly.
How to configure it
1. Set up your instances
You’ll need 2 instances:
- A hub instance — where APF itself is running and where you’ll orchestrate from.
- A remote instance — the external Pega application whose agents you want to expose and call into.
Use either the Pega 26.1 GA environment or the Infinity Agentic Experience instance;
Make sure your operator id in the hub instance has the PAPF access group. This will give you access to the Agentic Process Fabric Application
2. Getting the Remote Application Ready
- Create a application agent (if not here already)
- Make the agent ready for external access
- Create an OAuth 2.0 client credentials (client ID, secret)
For the secret (if not available) regenerate and copy the secret.
If you do not have the Client Secret
3. Register the remote application (From the Hub instance)
From the hub instance’s APF hub:
- Register the remote app using its OAuth 2.0 client credentials (client ID, secret, and the remote instance’s URL).
- Confirm the registration completes successfully — you should see the remote application listed as connected.
- At this point you should also be able to see the remote agents exposed by that application in the APF hub UI. If you can see them, registration itself worked
Use the details from your Remote App to register
- Use the application url (found in the Application Definition)
- Use Client ID and Client Secret (from step 2)
A registered App should look like this
3. Create or select the agents you want to orchestrate
This is where it’s easy to assume you’re done — but there’s a subtlety worth calling out:
Gotcha: If you create new agents after registering the application, you must go back and update those agents from the registered app.
Registration syncs the workflows available at the time it ran. If you add new agents to the remote application afterward, the hub’s picture of that app is now stale — the newly created agents won’t have their workflows populated on the hub side, even though the app registration itself still shows as healthy and the agents are visible.
Fix: re-sync/update the agents from the registered application. This repopulates the workflows into the agents on the hub, and remote case creation (or whatever action the agent performs) starts working correctly.
Symptoms if you skip this step: the remote app shows as registered, you can see the remote agents in the list, but when the Fabric Agent tries to actually create a remote case (or invoke the remote flow), it fails or silently does nothing — even though nothing in the registration itself looks broken.
Once done — go to the agent and check if the workflow menu shows the remote cases
4. Trigger the remote case creation (Success !!!)
Once the agents are updated:
- Invoke the Process Fabric Agent’s action to create the remote case (e.g., a Due Diligence case, a Claims case, whatever your remote app exposes).
- Verify the case is created on the remote instance and that the Fabric Agent correctly receives a reference/status back.
Quick troubleshooting checklist
If a Fabric Agent isn’t creating a remote case even though registration succeeded:
Is the remote app registration still active/healthy?- Are the remote agents visible in the hub?
- Were any agents created or modified after the initial app registration? If yes — re-sync/update them from the registered app.
- Do the agent’s workflows show as populated (not empty) after the sync?