NewCompare models in chat agent tests.See the benchmark

Automated testing for chat agents.

Voxli simulates customer and employee conversations to catch failures before launch and regressions when you change your agent.

Sign up See pricing

Works with any model your agent runs on

  • OpenAI
  • DeepSeek
  • Anthropic
  • Alibaba
  • Zhipu
  • Google
  • Mistral
  • Amazon
“We had our development team spend three months building internal tools to handle what Voxli does out-of-the-box.”
Senior Exec, Professional Services · Leading CX agent provider

Voxli's founders built ftrack and grew it from zero to millions in yearly revenue before selling it. ftrack won an Academy Award in 2021. Meet the founders

We build Voxli with our design partners

  • Kindly AI customer service agents
  • Expertise.ai AI sales agents
Read the Expertise.ai story

Before Voxli

A single answer doesn't show the whole conversation.

After a model, prompt, or tool change, check how your agent handles follow-up questions, corrections, and requests to act.

Facts.

It promises a delivery date that your order system never returned.

Memory.

Five messages in, it has forgotten what the user asked for.

Actions.

The customer only asks about a refund. It starts one anyway.

Tone.

Your brand's warm voice turns stiff and formal.

Test conversations from start to finish.

Define what your agent should do, then test it with simulated users who ask questions, change their minds, and push back.

Connect your agent.

Connect your chat widget or use a developer integration. A founder can help with setup.

Build your test set.

Write tests in plain English, or ask Voxli AI to explore your agent and create them from a prompt.

Simulate your users.

Run whole conversations with different personalities and languages to test how your agent handles varied requests.

See why a test failed.

Review the failed checks and the messages behind them to understand what went wrong.

Compare agent versions before you ship.

Run the same tests repeatedly on both versions. Compare shows what improved or regressed after a model, prompt, or tool change.

How to test a model swap

Run regression tests after release.

Rerun your suite after changes. Schedule simulated conversations against your agent in production, with Slack or email alerts when checks fail.

Use cases

How teams use Voxli.

Test your agent

Run regression tests and verify fixes before release.

Compare models

Choose the right model for your agent

Launch client agents

Test client requirements and review failures before sign-off.

Voxli Embedded

Offer agent testing inside your platform so clients can test their own agents.

We put 17 models behind one support and sales agent.

We used each vendor's default settings without tuning. Pick models from the benchmark, then test them on your own agent.

  1. 1 GPT-6 Luna OpenAI 87 Quality score
  2. 2 DeepSeek V4.1 Flash DeepSeek 85 Quality score
  3. 3 GPT-5.6 Luna OpenAI 85 Quality score
See the full benchmark

For developers

Investigate failures with your coding assistant.

Connect an MCP client to investigate failures and rerun tests.

More on MCP

Connect your workspace

claude mcp add voxli --transport http https://api.voxli.io/mcp

Then ask your assistant

Look up the latest run in Voxli and analyze why it failed.

  • Turn a bug report into a new test
  • Add checks to existing tests
  • Compare two runs side by side
Mattias LagergrenBjörn Rydahl

Start testing your agent.

Sign up for a 14-day trial. Mattias or Björn can help you get started.

Sign up Book a call

Try it free for 14 days.

Test your own agent with 1,000 trial credits. You don't need a credit card, and a founder can help you set it up.

See pricing