Skip to main content
A chat or text agent is an agent with Mode set to Text. Bluejay tests it by having a Digital Human exchange messages with your agent. There is no audio. Bluejay grades the conversation from the transcript against your success criteria and custom metrics.

​
How it works

  1. When a simulation runs, the Digital Human opens the conversation.
  2. Your agent replies.
  3. Bluejay threads each exchange into one conversation and keeps going until it ends.
  4. When the conversation ends, Bluejay grades the run against your success criteria.

​
Ways to connect a text agent

There are two main ways to connect a text agent. For the HTTP webhook path, pass the same simulation_result_id on every reply so it lands in the right conversation. See Chat Testing for the exact request and response shape. Some providers also expose their own chat connection types, for example Amelia, Google CES, Bland, and Kore. These are supported too.

​
Create a text agent

From the Agents view, click the + to add an agent. This opens the Connect Your Agent dialog. Fill it in:
  1. Name the agent.
  2. Write its system prompt.
  3. Set Agent type to Inbound or Outbound.
  4. Set Mode to Text.
  5. Pick the connection type, either SMS or HTTP webhook.
  6. Enter its details. For SMS this is a phone number. For HTTP webhook this is your webhook URL.
  7. Click Create.

​
Run a text simulation

Build a simulation on the text agent, add Digital Humans, and run it. Bluejay sends each Digital Human’s opening message, drives the back-and-forth, and grades the result.

​
What Bluejay measures

On a text run Bluejay grades your goals and custom metrics from the transcript. It does not produce voice quality signals and there is no recording.

​
Text uptime monitors

An uptime monitor runs the same message exchange on a schedule against a live text agent. On each run it checks that your agent responds and does not malfunction, so you catch a broken channel before your customers do. See Uptime Monitoring for scheduling, alerting, and verdict rules.

​
Set it up with the API or MCP

Text agents and runs are fully scriptable.
  • Create the agent with add-agent: mode="TEXT" and connection_type="SMS" with the phone number, or connection_type="HTTP_WEBHOOK" with the webhook URL. Over MCP this is the add_agent tool with the same fields.
  • Queue a run with queue-sms-simulation-run for an SMS agent, or queue-http-text-simulation-run for an HTTP webhook agent. Over MCP the queue_simulation_run tool covers both. Pass channel="sms" or channel="http_text".
Creating the simulation, adding Digital Humans, and reading results use the same tools as any other medium.