Ballet
Ballet turns a plain-English description of a multi-system business workflow into reviewable, version-controlled code that it then runs and maintains, writing integrations on demand against any API, including the internal and legacy systems no connector library covers.
What is Ballet?
Ballet is a workflow-automation platform built by the team behind Brainfish. It addresses a familiar deadlock: growth ideas stall the moment they require several systems to be wired together, and the engineering roadmap is never fast enough. The premise is to remove the queue rather than shorten it. You describe the outcome you want, whether that is lead routing, attribution or a renewal play, and Ballet builds it, runs it and keeps it working.
What separates it from connector-based tools is where those tools stop. Connector libraries cap out when a stack is proprietary or when the graph grows too large; Ballet writes the integration against any API at the moment it is needed, including internal ERPs and legacy tooling. The output is not a visual graph but version-controlled code, inspectable and reviewed diff by diff like a pull request. Execution is deterministic where accuracy matters and agentic only on the steps that genuinely benefit from judgment, which is the line the vendor draws against general-purpose task agents whose results vary between runs.
The homepage works through three examples: instant lead capture into a CRM, enriched and deduplicated before routing; account surge detection from product telemetry pushed to the right rep; and order changes propagated through internal billing without an engineering ticket. Around thirty-six third-party systems are named across the site, from Salesforce, Stripe and Slack to SAP, NetSuite and Okta.
Operational controls are prominent: evals on every step, auto-patching presented as a diff when an API changes, first-class approval gates, replayable runs, exportable audit logs, simulation mode, scoped permissions and snapshot rollback before anything writes to production. SSO and SAML sit at the Enterprise tier. Ballet is positioned as the execution layer of the Brainfish group, where Brainfish is the knowledge layer; workflows can be triggered from Brainfish or run standalone, and Ballet inherits Brainfish's SOC 2 Type II coverage. As of August 2026 the product is in early access, onboarding a small group of design partners, with a first workflow promised in a thirty-minute working session.
What it does
- Describe a multi-system workflow in plain English and have it built, run and kept working
- Generate an integration on demand, in seconds, against any system that exposes an API
- Run deterministic code on the steps that must be exact, with agentic reasoning only where judgment helps
- Catch breakage with evals on every step and receive the fix as a reviewable diff
- Gate every workflow behind code review and formal approval before it ships
- Simulate a run, scope permissions and roll back from a snapshot before anything writes to the CRM
- Replay any past run and export audit logs for review
When to use Ballet / When not to
A quick filter to help you decide if Ballet is the right fit.
When to use Ballet
- RevOps, MarketingOps and SalesOps teams whose cross-system revenue workflows keep stalling in the engineering queue
- Support Ops and CS Ops teams gluing ticketing, CRM and billing tools together by hand
- Engineering leaders who want the integration backlog cleared without pulling developers off the product roadmap
- Companies whose stack includes proprietary, internal or legacy systems that no off-the-shelf connector supports
- Teams automating production CRM, billing or identity systems that require formal review and approval before anything ships
When not to use Ballet
- Teams needing only simple single-trigger automations, which the vendor itself suggests keeping on Zapier
- Organisations whose target systems expose no API at all, a case the vendor places explicitly outside its scope
- Buyers who want self-serve signup, published pricing and an instant free trial
- Workloads governed by HIPAA, FISMA or GLBA, which the terms of service exclude from the service
- Teams with no engineering capacity to read and approve the generated code before it goes live
How to use Ballet
A typical end-to-end flow, from setup to results.
- Request early access from the signup page and create an account
- Wait to be selected, since onboarding is limited to a small group of design partners
- Book the thirty-minute working session offered to scope one workflow from your backlog
- Bring a real multi-system workflow and let the team model it on your actual systems
- Describe the outcome you want in plain English rather than drawing a flow by hand
- Let Ballet generate the integration code against each target system's API
- Review the generated code as a diff and approve it, the way you would a pull request
- Run it in simulation mode with scoped permissions before anything writes to production
- Approve the go-live, keeping the snapshot and rollback available
- Watch the evals, accept the auto-patch diffs when an API changes, and replay runs when you need to audit one
Pros & Cons
Pros
- Reaches the proprietary, internal and legacy systems that no connector library covers
- The generated code belongs to the customer: readable, version-controlled and forkable
- Deterministic on the steps that must be exact, unlike task agents whose output varies between runs
- Evals on every step catch breakage before it lands, and the fix arrives as a diff
- Serious guardrails before production writes: simulation, scoped permissions and snapshot rollback
- Cost is presented as predictable, indexed on workflows delivered rather than on task volume
- Solid inherited compliance base: SOC 2 Type II, ISO 27001 v2022, a public Trust Center, a published subprocessor list and no training on customer data
Cons
- Early access only: onboarding is restricted to selected design partners, with no general availability
- No public pricing, no published tier amounts and no plan grid
- Neither a permanent free plan nor a free trial is announced
- Target systems must expose an API, and Ballet itself publishes no documented API
- Web only, with no mobile application, and an English-only interface and documentation
- Terms and privacy policy are Brainfish's and cover the publisher's whole range rather than Ballet specifically
- Almost no public track record: the domain was registered in April 2026 and the product announced as launched on 11 August 2026
Pricing & Plans
No permanent free plan and no free trial are announced, and no price is published. The vendor states that pricing tracks the workflows delivered and their complexity, explicitly not per task, per recipe or per seat, and situates the entry point at the cost of a single forward-deployed engineer-month without converting that into a figure. Design-partner pricing is described as locked in before general availability, and an Enterprise tier exists, carrying SSO and SAML. The terms of service specify that all payments are made in US dollars unless otherwise contracted, by Mastercard, Visa or bank transfer, that purchases are non-refundable, and that a subscription may be cancelled at any time with effect at the end of the contracted term.
- pricing scoped to the workflows delivered and their complexity
- locked in before general availability
- no amount published
- tier carrying SSO and SAML
- no amount published
Data, GDPR & hosting
A consolidated view of how Ballet handles your data.
GDPR overview
GDPR implementation is concrete and documented. The privacy policy, last updated on 18 September 2025, covers the EU GDPR, the UK GDPR, the California CCPA and the Australian Privacy Act, and names Article 27 representatives: Prighter Group for the European Economic Area and Prighter Ltd for the United Kingdom, both reachable through a dedicated portal. Data subject rights are listed explicitly, including access, rectification, erasure, objection, restriction, portability, withdrawal of consent and marketing opt-out, and are exercised at a single privacy address. International transfers are framed by appropriate safeguards, with the EU-U.S. Data Privacy Framework and its UK and Swiss extensions referenced. No special category data is collected, a Data Processing Agreement governs customer content, and the publisher's site carries a GDPR compliance badge alongside EU data residency.
Who owns the data?
Under the terms of service, customers keep what they put in: the publisher asserts no ownership over user Contributions and states that all associated intellectual property rights remain with the customer. Generated workflow code is framed the same way, as version-controlled code the customer can read, fork and keep, with nothing shipping without sign-off. One exception is spelled out: Submissions sent directly to the vendor, such as questions, suggestions or feedback, have their intellectual property rights assigned to the publisher. For personal data, Brainfish acts as controller for account and website data, and as a processor bound by its Data Processing Agreement and the customer's instructions for customer content.
Reuse rights
The terms grant the end user a non-exclusive, non-transferable and revocable licence to access the Services and to download or print portions of their content, for internal business purposes only. Beyond that, nothing may be copied, republished, redistributed, sold, licensed or otherwise exploited commercially without prior written permission, and any wider use has to be requested from the publisher by email. Material coming from the customer side is treated differently: the vendor claims no ownership of Contributions, and the generated workflow code is described as the customer's to read, fork and keep. Reuse of personal data is governed separately by the privacy policy and the Data Processing Agreement.
Data retention & training
Hosting summary
The terms of service state that the Services are hosted in Australia, the United States and Germany, and that using them from another region means consenting to that transfer. The privacy policy is more specific about production: infrastructure sits primarily in the United States with AWS and Google Cloud, inside secure virtual private clouds with network segregation and firewalls. The publisher's own site advertises a choice of US, EU or AU data residency with data staying in-region, although that claim is made for the Brainfish platform rather than for Ballet specifically. The Trust Center publishes a subprocessor list in which AWS, GCP, Vercel, Datadog, ClickHouse and Qdrant are shown and six more are not, every disclosed subprocessor being located in the United States. International transfers are framed by appropriate safeguards, with the EU-U.S. Data Privacy Framework referenced. Buyers with strict residency requirements should confirm in writing which region a Ballet deployment actually runs in, since the two legal documents describe overlapping but not identical footprints.
Things to keep in mind
Risks and trade-offs to weigh before adopting Ballet.
- Ballet writes into production systems such as CRM, billing and identity, so a workflow approved too quickly can propagate a single error across five systems at once
- Approval gates, simulation mode and rollback only protect you if they are actually used, and they become easy to skip once the tool has earned a little trust
- Generated code still has to be read. A team that stops reviewing diffs because everything has always worked inherits a codebase nobody understands
- Removing the engineering queue also removes the friction that used to force teams to ask whether a workflow was worth building at all
- Operational knowledge risks migrating into generated workflows rather than into people, so when an automation finally breaks the person who understood the process may have moved on
- Pricing is not public, so budget exposure cannot be assessed before entering a sales conversation
- Terms and privacy policy cover the publisher's entire range, and data is hosted across Australia, the United States and Germany, which should be checked against your own residency commitments
Setup & Integrations
Technical difficulty
Low on the business side, moderate on the technical side. Describing a workflow requires no development skill and no engineering sprint, and the vendor builds the first one with you in a thirty-minute session. The technical prerequisites are real, however: every target system must expose an API and be granted access, and the generated code has to be read and approved by your own engineering team before it ships. Most design-partner workflows are announced as delivered within two weeks, the integration being generated in minutes while the bulk of the time goes to your scoping and approval.
Deployment
Integrations
Supported languages
Behind Ballet
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Ballet.
Frequently asked questions
What exactly is Ballet?
How is it different from Zapier or n8n?
Is the output deterministic?
Do our engineers keep control of the code?
Will it work with our internal or legacy systems?
Is it safe to connect it to our CRM?
How long does a workflow take to ship?
What does it cost?
How do security and compliance work?
Where does Brainfish fit in?
Should you pick Ballet?
Ballet is a serious attempt at a real problem. Ops teams everywhere are doing engineering work without engineers, and the tools sold to them cap out precisely where the work gets hard: proprietary systems, five-system chains, steps that must never drift. Writing the integration as version-controlled code, on demand, against whatever API exists is a credible answer, and the surrounding controls read as though someone has actually shipped automation into a production CRM and been burned by it: evals on every step, approval gates, simulation, scoped permissions, snapshot rollback and replayable runs.
The caution is commercial rather than technical. As of August 2026 Ballet is in early access with a hand-picked set of design partners, publishes no pricing at all, and offers neither a free plan nor a trial. Its legal terms, privacy policy, certifications and social presence all belong to Brainfish, because Ballet is a product rather than a company. The domain was registered in April 2026 and the product announced as launched on 11 August 2026, so there is no public track record to weigh, and the claim that it generates working integrations against real internal systems stays a claim until someone outside the design-partner group can test it.
What tips the balance is the publisher behind it. Brainfish has raised USD 10 million, holds SOC 2 Type II and ISO 27001 certification, publishes a subprocessor list and states that it does not train on customer data. That is an unusually solid foundation for a product this young. The reasonable position today is to treat Ballet as a design-partner engagement rather than a purchase: if one multi-system workflow has been stuck in your backlog for two quarters, the thirty-minute working session costs nothing and will tell you more than any comparison table.
- Choosing a selection results in a full page refresh.
- Opens in a new window.