Metatable
Metatable turns a plain-English description of a business process into a complete, deployed application. It writes a readable plan you approve first, then generates the database, backend, screens and hosting, with full source-code export included.
What is Metatable?
Metatable is a spec-driven software generation platform aimed at people who know their business but cannot safely judge code. Instead of taking a prompt and improvising an application, it first writes down how the software should work, covering roles, records, rules and edge cases, as a plain-language product requirements document. You read that document, correct it and approve it; only then is anything built. The same plan stays the source of truth afterwards, so changing the business means changing the plan rather than patching a codebase nobody can audit.
What it produces is a complete, deployed system rather than a prototype. Every generated application ships with a real PostgreSQL database with proper schema and relations, a compiled Rust backend exposing typed API endpoints, modules and scheduled jobs, a Flutter front end serving web and mobile from one source, and built-in authentication by email or Google with the permissions the plan described. Secrets for third-party APIs sit in an encrypted vault. Hosting on managed Azure infrastructure starts the moment the build finishes, on a provided sub-domain or a custom one.
Two commitments frame the offer. The first is ownership: the full source code, frontend, backend and database, can be exported at any time on every plan, and the company states there is no lock-in. The second is headless use: the generated backend can expose an MCP server, so ChatGPT or Claude can query the application or trigger workflows without any dashboard.
The company publishes unusually candid numbers. The homepage claims 4,000+ registered users, 5,000+ projects built and a typical first version in about thirty minutes. The features page reports 87% of backend builds succeeding over ninety days, and the about page 95% database generation success against 72% backend deployment on an October 2025 cohort of eighteen builds. Metatable also states plainly where it is weak: front-end generation is still maturing, and it is not the right tool for games or products whose interface is the point. Its declared sweet spot is CRMs, inventory, portals, trackers, approval flows and operational dashboards.
What it does
- Turn a plain-English description of a business process into a working application
- Produce a readable specification you approve before any code is generated
- Generate a PostgreSQL database, a Rust backend and a Flutter front end from that specification
- Deploy and host the finished application on managed Azure infrastructure
- Expose the application as an MCP server usable from ChatGPT or Claude
- Export the complete source code at any time and run it elsewhere
- Update the plan when the business changes and let the software follow
When to use Metatable / When not to
A quick filter to help you decide if Metatable is the right fit.
When to use Metatable
- Small business owners running operations on spreadsheets who need custom internal software without hiring developers
- Operations managers replacing manual inventory, purchase-order and approval workflows with a real application
- Freelancers and small agencies delivering deployed client software with readable specs and exportable source code
- Procurement and contract teams building vendor registries that track terms, renewals and spend
- Growing teams needing internal tools without pulling engineers off the core product
When not to use Metatable
- Teams whose product is the interface itself, built on rich animations, custom visuals and design polish
- Game developers and builders of highly customised consumer-facing applications
- Individuals wanting a personal or hobby app, since the Terms restrict use to companies and sole traders
- Organisations requiring a signed data processing agreement or a security certification such as SOC 2 or ISO 27001
- Non-English-speaking users, as the platform, the specifications and the documentation are English-only
How to use Metatable
A typical end-to-end flow, from setup to results.
- Create a free account on the Metatable app, no credit card required
- Describe the business process in plain English: workflow, records, rules, roles and edge cases
- Answer the follow-up questions about your business, or accept the answers Metatable suggests
- Read the generated product requirements document and correct anything that misstates the business
- Approve the plan, since nothing is built before you do
- Wait while the database, backend, front end and tests are generated and validated
- Review the deployed application on its Metatable sub-domain, then publish it
- Add a custom domain and invite editors or team members according to your plan
- Connect the backend as an MCP server to use the app from ChatGPT or Claude
- Update the plan whenever the business changes, or export the full source and host it yourself
Pros & Cons
Pros
- A readable plan is approved before generation, so a non-technical owner can verify what will be built
- Delivers a complete deployed system with database, logic, screens, logins and hosting, not a prototype
- Full source-code export on every plan, with ownership of Output written into the Terms
- No technical setup: no stack to choose, no cloud account to wire, no server to maintain
- Permanent free plan with no credit card, and a low paid entry point at 25 USD per month
- Unusual candour about measured success rates and about where the product is still weak
- A 99.0% uptime SLA backed by service credits, and a named list of fifteen sub-processors
Cons
- Front-end generation is still maturing by the vendor's own admission, and visual polish is not its strength
- Backend deployment is not guaranteed: 87% success over ninety days, and 72% on an earlier cohort
- Services are supplied as is with no warranty, and liability is capped at 70% of the last annual fees
- Payments are non-cancellable and non-refundable, with no proration and no rollover of unused quota
- No data processing agreement, no security certification and no documented AI-training opt-out
- Use is contractually restricted to companies and sole traders, and the interface is English-only
- Free-plan backends shut down every four hours when idle and come with no backups
Pricing & Plans
Metatable operates on a freemium basis. A permanent free plan is available without a credit card and covers a single project. The lowest paid tier, Standard, is priced at 25.00 USD per month; Pro is 79.00 USD per month and Enterprise starts at 700.00 USD per month. Subscriptions may be taken monthly or annually, are payable in US dollars and must be prepaid in full. Usage beyond a plan's quota of AI tokens and HTTP requests is billed separately, either at the end of the billing period or once it reaches 5 USD, whichever comes first. Payments are non-refundable.
- one owner
- 1 project
- 250k AI tokens per month
- 1 GB storage
- 1k requests per month
- requirements chat
- structured PRD and app structure review
- no backups
- one owner plus up to two editors
- 5 projects
- 1M AI tokens per month
- 10 GB storage
- 100k requests per month
- adds always-on hosting
- MCP server
- project export and 90-day backups
- up to 10 total users
- 25 projects
- 4M AI tokens per month
- 50 GB storage
- 500k requests per month
- adds a premium shared instance
- more room for team review and 365-day backups
- custom team setup
- unlimited projects
- unlimited AI tokens
- 200 GB storage
- unlimited requests
- adds a dedicated instance
- human-assisted development and custom architecture discussion
Data, GDPR & hosting
A consolidated view of how Metatable handles your data.
GDPR overview
Implementation is documented and specific rather than merely declared. The privacy notice, updated 1 January 2025, sets out GDPR and UK GDPR legal bases and adds regional regimes for Switzerland, Canada and Brazil. It lists the full set of data subject rights, points to the EDPB member authorities, the UK ICO and the Swiss FDPIC for complaints, and commits to notifying the supervisory authority within 72 hours of a breach. Transfers out of the EEA rely on Standard Contractual Clauses or the EU-US Data Privacy Framework, to which Metatable declares adherence including the UK and Swiss extensions. An annex names fifteen sub-processors with their processing location, and the stated contact is a Data Protection Specialist reachable by email. No EU representative under Article 27 is named, and no data processing agreement is published or offered.
Who owns the data?
Under the Terms of Service the customer retains all ownership rights in the Input supplied to Metatable and owns all Output, with Metatable assigning to the customer whatever rights it might hold in that Output. The generated source code can therefore be exported and run anywhere, with no lock-in. Metatable itself keeps ownership of the platform and grants only a right of use. Legally it acts strictly as a data intermediary and processor, or sub-processor, which makes the customer the controller. One reservation: clause 7.2 lets Metatable store, aggregate and use data generated while performing the Services for the legitimate purposes of those Services.
Reuse rights
Yes. Because the customer owns all Output, the generated code, data and documents can be reused, modified, exported and run on any infrastructure without asking Metatable for permission, and source export is offered on every plan as a core part of the product. Two contractual limits apply: Output may not be used to develop services that compete with Metatable's own products, and data may not be extracted from the platform by methods other than those the Terms and platform policies allow. On Metatable's side, prompts and outputs may be sent to third-party AI providers such as OpenAI, Anthropic, LlamaIndex and Groq for processing, and the company may store and aggregate data generated by the service in order to operate it. Nothing in the published documents states that customer data is used to train AI models.
Data retention & training
Hosting summary
Metatable states that its primary servers and data centres are located in the United States. Generated applications run on managed Microsoft Azure infrastructure, and the sub-processor annex documents Google Cloud Platform providing cloud computing and hosting from region-based data centres across the US and the EU, with Azure holding backups on the same US/EU basis. The AI providers that process prompts and outputs, namely OpenAI, Anthropic, LlamaIndex and Groq, are all listed as processing in the USA, as are Stripe, R! B2B, SendGrid and Intercom; Snitcher processes in the EU, and GitHub and GitLab across the US and the EU. Transfers out of the EEA, the UK or Switzerland rely on Standard Contractual Clauses or the EU-US Data Privacy Framework, to which Metatable declares adherence including the UK and Swiss extensions. Stated safeguards include TLS/SSL encryption in transit and at rest, strict access controls, firewalls and intrusion detection, regular audits and penetration testing, and mandatory non-disclosure agreements. Governing law is that of Delaware. No security certification such as SOC 2 or ISO 27001 is mentioned anywhere.
Things to keep in mind
Risks and trade-offs to weigh before adopting Metatable.
- Prompts and generated outputs are routed to third-party AI providers such as OpenAI, Anthropic, LlamaIndex and Groq, so anything typed into a specification leaves Metatable's own infrastructure
- Approving a plan you skimmed rather than read shifts responsibility for the result onto you, and a misstated rule becomes a bug that runs unattended at 3am
- Running a business on generated software you cannot review yourself creates a dependency you may only notice the day something breaks
- The service is supplied as is with liability capped at 70% of annual fees, so an operational failure is largely your loss rather than the vendor's
- Payments are non-refundable with no proration, and Metatable may unilaterally change or deprecate features, leaving only 30 days to react
- The non-solicitation clause carries a 30,000 USD penalty per violation for two years after termination, and breaches may be disclosed publicly
- No documented way to exclude your data from AI training, no DPA and no security certification, which makes it a poor fit for regulated or sensitive data
Setup & Integrations
Technical difficulty
Very low. Metatable advertises no coding and no developers: there is no stack to choose, no cloud account to wire and no server to maintain. Sign-up needs no credit card, and the entry point is a conversation in plain English rather than a technical form, since the questions asked are about customers, stock and rules rather than frameworks. A first working version is claimed in about thirty minutes. The one step demanding real attention is reading and approving the specification. Difficulty only reappears if you export the source and host it yourself. Interface and documentation are English-only.
Deployment
Integrations
Supported languages
Behind Metatable
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Metatable.
Frequently asked questions
Who owns the code Metatable generates?
Is Metatable a no-code builder?
Do I need a credit card to start?
How long does a first working version take?
What technology does it generate?
Can I use my application from ChatGPT or Claude?
What happens if I cancel?
Where is my data hosted?
Is a data processing agreement available?
What is Metatable not good at?
Should you pick Metatable?
Metatable answers a real and specific problem: the gap between an owner who knows exactly how their business works and the software that would run it. Its bet is the approved plan. By writing the specification first, in language a non-technical reader can check, it makes the output verifiable by someone who could never audit a codebase, and that is a more honest guarantee than most prompt-based builders offer. What arrives is genuinely complete: a PostgreSQL database, a compiled Rust backend, a Flutter front end, authentication, an encrypted secrets vault and managed Azure hosting. Ownership is written into the Terms rather than merely advertised, and full source export on every plan makes an exit possible. The permanent free plan removes the risk of trying it. The reservations are real and, to the company's credit, mostly stated by the company itself. Front-end generation is still maturing, and backend deployment succeeds 87% of the time over ninety days rather than always. Contractually the service is supplied as is, liability is capped at 70% of the last annual fees, payments are non-refundable, and Metatable may unilaterally change the scope of a subscription. The compliance picture is documented but incomplete: GDPR bases, subject rights and a named sub-processor list are published, yet there is no DPA, no security certification and no documented way to exclude your data from AI training, while prompts and outputs are routed to OpenAI, Anthropic, LlamaIndex and Groq. Two structural points deserve attention: the company is young and thinly funded, and its legal documents carry two different Delaware addresses. Metatable is worth the free plan for any operator whose process is documented and whose priority is a working system rather than a beautiful one. Read the Terms before committing to a paid tier.
- Choosing a selection results in a full page refresh.
- Opens in a new window.