Skip to main content
Bluejay connects directly to your Google Conversational Engine (CES) agents so you can test them with Digital Humans over Voice or Chat — sync your agents, connect one, and run simulations against it.

​
Prerequisites

  • A Google Cloud project with the Conversational Engine API (ces.googleapis.com) enabled.
  • At least one CES app in that project.
  • Permission to grant a service identity access to the project (to complete the steps below).
Enable the API from APIs & Services → Library — it should read API Enabled: Conversational Engine API enabled in the Google Cloud console
CES shares the same “Google Agents” integration and the same service-account key as Dialogflow CX — one credential covers both Google agent types. If you have already set up the shared credential for DFCX, you only need the CES-specific permissions and the steps below. See the Dialogflow CX page for the shared credential setup.

​
1. Grant Bluejay access (service account)

Bluejay authenticates to your Google Cloud project with a service-account key that you create and paste into Bluejay. The key is encrypted at rest and only ever used to call the Google APIs for your agents.

​
Step 1 — Create a service account with the right permissions

In the Google Cloud Console → IAM & Admin → Service Accounts:
  1. Create Service Account → give it a name (e.g. bluejay-ces).
Create the service account
  1. Grant the minimal CES permissions. CES is read + run only — there is no edit/push tier. CES apps live under the Conversational Engine API (ces.googleapis.com). Create a custom IAM role (IAM & Admin → Roles → Create Role) with exactly the permissions below, or use the pre-built role listed for each group.
roles/ces.viewer and roles/ces.client are complementary, not overlapping — viewer has the app/deployment reads but no session permission; client has the session permissions but no reads. So the pre-built path is roles/ces.viewer + roles/ces.client + roles/serviceusage.serviceUsageConsumer, or roles/ces.admin (covers read + run) + roles/serviceusage.serviceUsageConsumer.
serviceusage.services.use is required regardless of which path you pick — without it, runtime calls fail immediately with USER_PROJECT_DENIED (Bluejay sends an x-goog-user-project quota header on every request).

​
Step 2 — Create a JSON key

On that service account, open the Keys tab → Add Key → Create new key → JSON → Create. A .json key file downloads to your machine.
Treat the key file like a password — it grants the permissions above on your project until you rotate or delete it. You can revoke it anytime from the same Keys tab.

​
Step 3 — Add it to Bluejay

  1. Log into your Bluejay Dashboard.
  2. Click your Organization menu (bottom-left) → Integrations → Google Agents.
  3. Paste the contents of the JSON key file into Service Account Key (JSON).
  4. Click Save.
Paste the service-account key in the Google Agents integration This is the same credential used for Dialogflow CX — a single key covers both Google agent types. The key is encrypted at rest and only ever used to call the Google APIs for your agents.

​
2. Sync your Google agents

Once the credential is saved, discover your agents:
  1. In Settings → Integrations → Google Agents, click Sync.
  2. Bluejay scans your project across all locations and creates an agent in Bluejay for every CES engine it finds (alongside any Dialogflow CX agents).
The sync result reports how many agents were created, updated, or removed (an agent deleted in Google is flagged in Bluejay rather than dropped). Re-run Sync any time you add agents in Google.

​
3. Connect an agent

Most agents are wired up automatically by Sync. To connect or adjust one manually:
  1. Open the agent’s Connection settings.
  2. Choose the Google Conversational Engine (CES) connection type — Voice or Chat.
  3. Set the CES connection fields:
That’s all that’s needed — the service-account key from Step 1 handles authentication.

​
4. Test your agent

With an agent connected you can run simulations against it like any other Bluejay agent. In Simulations → Create Simulation, pick a simulation type and run Digital Humans against your CES agent: Create Simulation — pick a type CES supports both Voice and Chat, chosen by the agent’s connection type:
  • Voice simulations bridge audio between a Bluejay Digital Human (the caller) and your CES agent in real time.
  • Chat simulations drive a turn-by-turn text conversation.
Tool calls are handled automatically. When your CES agent invokes a tool or function call mid-conversation, Bluejay captures the invocation (tool name + arguments), returns a response so the turn keeps flowing, and sends that tool response back to CES — no setup required. Today the returned value is a default acknowledgment ({"status": "ok"}); configurable/scripted tool responses are on the way. Server-side tools that CES runs itself (data stores, OpenAPI/connectors) still execute normally on Google’s side; this only covers the client-side calls handed back to the caller.
Author your Digital Humans directly and run them against the connected agent.

​
Summary