# Functionize Studio — Agentic end-to-end test platform that builds, runs and repairs tests against your live web app.

> Source: The Agents Index — https://theagentsindex.com/functionize (structured, researched, re-verified)
> Facts last verified: 2026-10-24

Functionize Studio is an agentic end-to-end testing platform that turns a plain-language description into a test, runs it across browsers and environments, and repairs it when the app changes. It runs on Functionize's own element and execution models and connects to coding agents over MCP.

| Fact | Value |
| --- | --- |
| Website | https://functionize.com |
| Price | free, then from $20/month |
| Free tier | Yes |

$0 with no time limit: all agent capabilities, five parallel runs, a monthly credit grant (the plan card says 800,000 credits/mo, the FAQ says 400,000), team workspace, self-serve support only, no card required; no SSO, RBAC or add-on credit purchase.

## What it does

**Writes code** · Partly — It authors complete tests from a plain-language sentence, but the vendor positions coding agents as the writers of application code and Studio as the independent check on it.

> Agents write the code. ‍Studio proves it works.
> — https://www.functionize.com/ · Hero headline, second line

**Debugs and fixes** · Partly — Studio diagnoses a failing run, names the cause and lands the fix in the test itself; the repair is to the test step, not to the application source that broke it.

> Studio investigates the run and **names the failing step and the cause**. Then it applies the fix and validates it with a run.
> — https://www.functionize.com/agent-tasks · Put Studio to work, 'Diagnose and fix' card

**Writes tests** · Yes — Generating end-to-end tests from a described flow is the core job: steps, element identification and verifications come back generated and already run.

> The full test comes back: **steps, element identification, and verifications**, generated from your sentence and run immediately.
> — https://www.functionize.com/agent-tasks · Put Studio to work, 'Create a test' card

**Ships it** · No — It stops at proving the change: suites, reports and fixes come back, while the release decision is explicitly left with the buyer rather than automated.

> You still own what good means, which risks matter, and the call to ship
> — https://www.functionize.com/agent-tasks · Capabilities section, 'Studio runs all of it' closing line

**Done-for-you service** · No — This is software the buyer runs themselves: sign-up is self-serve, there is nothing to stand up, and no engagement where Functionize staff build or operate suites for you.

> Log in and start testing. No infrastructure, nothing to maintain.
> — https://www.functionize.com/architecture · How it works, 'Platform Interface' block

## How much it does unattended

**Runs autonomously** · Yes — A single prompt produces a built, executed and self-maintained test, including re-runs on failure and fixes applied without a person in the step.

> Describe any flow in plain language. Studio builds the test, runs it across browsers, and fixes it when the app changes.
> — https://www.functionize.com/agent-tasks · Hero paragraph

**Multi-agent** · Yes — The architecture splits the work between a planning agent and an execution agent, with the plan written so a second agent can act on each ordered step.

> An agent reads what you asked for and maps it into ordered test steps, each one specific enough for another agent to act on.
> — https://www.functionize.com/architecture · How it works, 'Agent Layer' block

**Agent permissions** · Yes — You set how far the agent may go and every edited step writes a restore point you can inspect or roll back with one click, so changes are reversible.

> Any edited step writes a restore point that records what changed, on which step, and why, and you roll back to the exact state before it with one click.
> — https://www.functionize.com/architecture · FAQ, 'What if the AI fixes something I didn't want fixed?'

## Which models it runs on

**Claude** · No — The reasoning and element models are Functionize's own rather than a hosted frontier model; Claude appears only as an MCP client that calls Studio, not as a model option.

> The models, the data layer, and the execution engine are ours. Built for testing, sharpened on your app.
> — https://www.functionize.com/ · Meet Studio, 'Our intelligence' block

**GPT** · No — The vendor answers the wrapper question directly: the element model and data layer are in-house and no third-party model endpoint sits under a result.

> No. The element model and the data layer are built in-house and trained on 200M+ UI datapoints from real runs.
> — https://www.functionize.com/architecture · FAQ, 'Is this a GPT wrapper?'

**Gemini** · No — No external model provider is selectable; the vendor states there is no moving third-party endpoint beneath a verdict, which rules Gemini out alongside the rest.

> there's no moving third-party endpoint underneath the result
> — https://www.functionize.com/architecture · FAQ, 'Is this a GPT wrapper?'

**Model choice** · Partly — The vendor picks the models by default, and only the enterprise deployment bullet offers bringing your own model and data alongside private deployment.

> ✓ Your model, your data
> — https://www.functionize.com/architecture · Enterprise ready section, deployment checklist

## Where you use it

**In your editor** · Yes — The MCP server puts test creation and diagnosis inside the editor-side agents developers already run, with config snippets for Cursor, Windsurf, Gemini and Codex.

> Build, run and diagnose tests against your real app from Claude Code, Cursor or Copilot
> — https://www.functionize.com/mcp · Hero subheading

**In your pipeline** · Partly — Enterprise prose claims CI/CD hookup through the same integration layer, yet the integrations page itself is a 'Coming Soon' placeholder with an MCP waitlist, so pipeline use is not demonstrably live.

> And it plugs into the tools your teams already use: your coding agents through MCP, CI/CD, Jira, Slack, and your identity provider.
> — https://www.functionize.com/enterprise · Enterprise apps section, intro paragraph

**In a browser** · Yes — Studio is a hosted web application: you log in, prompt it in chat, and read runs, reports and credit usage in its sidebar.

> You can see your usage in the Studio sidebar and in Reports.
> — https://www.functionize.com/pricing · FAQ, 'Where can I see my usage?'

## Whose machine it runs on

**Self-hosted** · Partly — Execution normally runs in the Functionize cloud, with deployment inside your own environment offered as part of the enterprise posture rather than on self-serve plans.

> Deploy in the Functionize cloud or inside your own environment
> — https://www.functionize.com/architecture · Enterprise ready section heading

**Open-source runtime** · No — There is no way to run it without Functionize: both the web product and the MCP server require signing in with a Functionize account before any agent work happens.

> Sign in with your Functionize account in the browser. No API key.
> — https://www.functionize.com/mcp · Connect section, step 3

**Runs in your cloud** · Partly — For internal or VPN-bound environments the execution layer runs inside your network over encrypted tunnels, but the default is the vendor's isolated cloud.

> When you need to reach internal or VPN bound environments, it runs inside your network through secure encrypted tunnels that keep everything inside your perimeter.
> — https://www.functionize.com/architecture · FAQ, 'Do I have to install or maintain any infrastructure?'

## Building with it

**Tracing and evals** · Yes — Every run is recorded with pass/fail, browser and duration, each step screenshotted, each element match shown with its bounding box, and reports export to CSV.

> Every run recorded: pass or fail, browser, duration. Filterable reports, exportable to CSV.
> — https://www.functionize.com/agent-tasks · Capabilities, 'Reporting and history' card

## What it costs to run

**How it meters** · Yes — Billing meters agent work in credits from one shared balance; human work such as editing, reading reports or changing settings consumes nothing.

> A credit is the unit that measures agent work. Studio subtracts every agent task from the same credit balance.
> — https://www.functionize.com/pricing · FAQ, 'What is a credit?'

**Price per 1,000 requests** · Yes — $0.005 per 1,000 requests in Scale, its $100 plan fee over 20,000,000 requests, its cheapest mode.

> $100/mo
> — https://www.functionize.com/pricing · Scale plan card, price cell

**Free credit** · Yes — 800,000 free credits, a month.

> 800,000 credits/mo
> — https://www.functionize.com/pricing · Free plan card, allowance line

**Free tier** · Yes — A standing $0 plan with no time limit and no card gives all agent capabilities, five parallel runs and a monthly credit grant.

> The Free plan includes all agent capabilities, 400,000 credits each month, and five parallel runs. You do not need a credit card.
> — https://www.functionize.com/pricing · FAQ, 'What does the Free plan include?'

**MCP server** · Yes — A hosted MCP server at mcp.functionize.com lets any MCP client build, run and diagnose tests in chat, with OAuth sign-in instead of an API key.

> The Functionize MCP server connects your coding agent to Functionize Studio. Ask for a test in plain language and the result comes back in the chat.
> — https://www.functionize.com/mcp · Hero paragraph under 'Functionize MCP'

**Bring your own key** · Partly — Inference is normally the vendor's own; bringing your own model and keeping your own data is presented only as part of the enterprise deployment option.

> ✓ Your model, your data
> — https://www.functionize.com/architecture · Enterprise ready section, deployment checklist

## Buying it for a team

**A company can buy it** · Yes — Enterprise buys on invoice or PO with a twelve-month minimum, multi-team management and premium support, alongside self-serve card plans for smaller teams.

> Invoice/PO billing, 12-month minimum
> — https://www.functionize.com/pricing · Enterprise plan card, bullet list

**Seat model** · Partly — Plans are priced per plan and credit allowance rather than per seat, but seats are still charged and prorated at a rate set in your order form or online agreement.

> New seats are charged at the agreed upon rate in your Order Form or online purchase agreement.
> — https://www.functionize.com/usage-limits · How credits work table, 'Adding Users mid-cycle' row

**Pooled budget** · Yes — Growth, Scale and Enterprise draw agent work from one team-wide credit balance; only individual allocations are described as unpooled.

> The Growth plan and the Scale plan use one credit balance for the whole team.
> — https://www.functionize.com/pricing · FAQ, 'How do credits work for a team?'

**Admin controls** · Partly — Role-based access control arrives with the cheapest paid plan and is absent from Free, while multi-team management is held back for Enterprise.

> SSO and Role-Based Access Control
> — https://www.functionize.com/pricing · Growth plan card, Includes bullet list

**Audit log** · Partly — A full audit trail is listed among the enterprise deployment controls; nothing on the self-serve plan rows offers administrative logging.

> ✓ Full audit trail
> — https://www.functionize.com/architecture · Enterprise ready section, deployment checklist

**Single sign-on** · Partly — SSO ships from the $20 Growth plan upward, and the plan comparison marks it absent on Free, so the free tier cannot be put behind an identity provider.

> SSO and Role-Based Access Control
> — https://www.functionize.com/pricing · Growth plan card, Includes bullet list

## What happens to your code

**Opt out of training** · Partly — Raw customer content is not used to train models for other customers without written consent, but de-identified aggregated usage metadata is used for model refinement with no stated opt-out.

> We expressly do not use raw Customer Content or proprietary application data to train models for the benefit of other customers without your explicit written consent.
> — https://www.functionize.com/privacy-policy · Section 2, Data Use and AI Training Refinement

**Getting out** · Partly — Exporting results is free and never costs credits, but unused credits expire at each reset and are forfeited on cancellation with no refund or proration.

> No. If you cancel, the subscription stops and you lose the credits that remain.
> — https://www.functionize.com/pricing · FAQ, 'Can I get a refund for credits?'

**Certifications** · Yes — SOC 2 Type II is claimed with the report available to a security team on request, and GDPR and CCPA data rights are set out in the privacy policy.

> Furthermore, we maintain a rigorous audit schedule and adhere to SOC 2 Type II compliance standards to verify the ongoing effectiveness of our security controls.
> — https://www.functionize.com/privacy-policy · Section 4, System Security and Integrity

## What changed

- 2026-10-24 — Capabilities · Free tier · Plan limits · Price · Terms

## Keeping it current

These facts are re-checked against Functionize Studio's own pages and updated when they change. If you represent Functionize Studio, sign in and update the listing from your dashboard at https://theagentsindex.com/dashboard. The operations this site accepts from an agent are listed at https://theagentsindex.com/.well-known/agents.json — that catalog needs no key to read, and each entry states its own auth.