Testzilla Console (TZ Console)

TZ Console is the AI assistant built into Testzilla, and the Console is its full-screen home in the app. It is the AI-native, easiest way in: describe what you want in plain language and TZ Console builds anything (projects, channels, tests) and connects anything (your integrations and providers), all in one place. No forms, no wizard hopping. If you are comfortable with AI, this is where to start.

Think of the Console as the no-code power surface: everything the API and MCP server can do, driven by conversation, scoped to your account. It works across the whole platform and across the systems you have connected through Integrations.

Just want a first result?

The fastest path is to ask TZ Console. From the Dashboard or Projects page, click Set up with TZ Console / Ask TZ Console and describe what you want to test. See the Quickstart for the no-code walkthrough.

What TZ Console can do

Build your test setup from a sentence

Describe the agent you want to test and TZ Console assembles the whole hierarchy for you — no forms:

• Creates a Project to group everything.
• Adds the right Channel for your agent and picks the connection type (phone, LLM Chat, Web Chat, or Web Voice).
• Writes the Tests — scenarios and pass/fail criteria — from your description.
• Runs them, and reports back the results.

“Test my Retell support line on +1 555 0100. Make sure it can book an appointment and never gives medical advice.”

From that one sentence TZ Console creates the project, the phone channel, the booking and safety tests, and starts the runs.

If the channel has an Agent System Prompt — the system prompt of the agent under test — TZ Console reads it as context when generating the suite, so the tests cover what the agent actually claims to do. The prompt is your IP: it is read-only context for TZ Console and is never edited.

Use your connected systems through Integrations

When you connect a system under Integrations (Retell, Vapi, ElevenLabs, and others), TZ Console can read from it directly — so you do not have to look up IDs and copy them in by hand. It can list the agents in your connected account, pull the agent and phone-number details a channel needs, and wire them onto the channel for you.

Manage schedules

Ask TZ Console to schedule runs instead of triggering them by hand — for example, run a regression suite nightly, or before each release. It sets up the scheduled runs and you track them on the Queue and Dashboard.

Wait on an external event (park and resume)

Some work cannot finish in a single turn — it depends on something happening outside Testzilla, like a webhook firing or a call completing. When that happens, TZ Console can park the turn instead of failing or spinning: it shows a waiting on external event banner with a short reason and, where relevant, the postback address the external system should call. When the event arrives, the Console resumes the same conversation automatically and continues from where it left off. You do not have to babysit it or re-ask — a parked turn simply picks back up when its event lands.

Improve your prompts from real results

This is where the Console earns its keep. TZ Console reads your actual test results — the transcripts, failures, and analyses — and proposes concrete improvements:

• Sharpen Testzilla’s own prompts — tighten a test’s scenario or pass/fail criteria, or the shared project and channel prompts that tests inherit.
• Fix the agent under test — when results reveal a weakness, TZ Console can suggest edits to the prompt of the connected agent itself, and (for systems connected through Integrations) help apply those changes back to the source agent.

The loop is: run tests -> read results -> improve prompts -> re-run. TZ Console drives all of it from chat.

Move and port agents between systems

Because TZ Console can read connected systems through Integrations, it can help move or port an agent’s configuration between Integration accounts or systems — for example, lifting an agent’s setup from one connected account to another, so you can stand up the same agent in a new environment and test it there.

Caution

TZ Console acts on your behalf and can create, change, and run things that cost credits, as well as propose edits to connected agents. Review what it suggests before applying changes to a production agent.

Console runs use the same run gate

Work in the Console that spends credits is subject to the same subscribe-first run gate as the rest of the app: if the account has no active subscription, or has reached its balance or overage limit, the Console stops before doing the paid work and asks you to subscribe or top up rather than running the balance negative. See subscriptions and the run gate.

Example prompts

TZ Console understands plain language, but every request maps to something the platform can actually do. The prompts below are grouped by what they drive. You can phrase them naturally; the short forms in brackets are the underlying commands.

Look around your account

• “List my projects.” (list projects)
• “Show the channels in this project.” (list channels)
• “List the tests on this channel.” (list tests)
• “Show my recent results for test abc123.” (results abc123)
• “What’s my credit balance?” (balance)
• “Who am I logged in as?” (whoami)

Focus the conversation

• “Use project abc123 as context.” (use project abc123)
• “Use channel xyz789.” (use channel xyz789)
• “What context am I in right now?” (context)
• “Clear the project context.” (clear project)

Build and run

• “Create a project for my support line, add a phone channel for my Retell agent on +1 555 0100, and write a booking test.”
• “Run test abc123.” (run test abc123)
• “Run batch xyz789.” (run batch xyz789)
• “How is run abc123 going?” (status abc123)
• “What’s in the queue?” (status queue)

Learn from results

• “Analyze my failures from the last 7 days.” (analyze failures —days=7)
• “Analyze test result abc123 and tell me why it failed.” (analyze test abc123)
• “Read my failing booking tests and suggest prompt improvements.”

Check connectivity

• “List the MCP tools available.” (mcp)
• “List the agents in my connected Retell account.” (uses your Integrations)

One test run is one run

Running a single test with run test runs it once. There is no iterations option on a single test run. To run the same tests N times for reliability, use a Batch: run batch <id> --iterations=<n> (1 to 50). Iterations are a batch feature, not a single-test one.

Where to find it

The Console lives in the left nav of the web app. A compact Ask TZ Console entry point also appears on the Dashboard and Projects pages for quick, in-context requests; the full-screen Console is for longer, multi-step work.

Console vs API vs MCP

All three drive the same Testzilla backend, scoped to your account — pick the surface that fits how you work:

Console

No code. Describe what you want in plain language inside the app. Best for setup, exploration, and improving prompts from results.

MCP

Drive Testzilla from your own AI agent (Claude Code, Cursor, and others) over the Model Context Protocol. See Integrate via MCP.

API

Automate from your own code and CI over REST. See Integrate via API.