The API nobody wants to write
Almost every internal API is the same API. A handful of routes over a handful of tables. Pagination, because returning everything is not an option. Filtering and sorting, because the caller never wants all of it. An API key check. Some documentation that is accurate for about a fortnight. Then error handling, logging, and a rate limit after the first time something hammers it.
None of it is difficult. All of it takes time, and none of it is the thing you were actually trying to build. Worse, it does not end at launch: every new column, table or consumer means going back into the code, and the documentation drifts a little further from the truth each time.
Dataloom generates that layer from your schema instead. You describe what should be exposed, and the routing, pagination, filtering, documentation, keys and logging come with it — and stay consistent, because they are derived rather than hand-written.
How the builder works
- 1
Connect a data source
Enter your connection details, or upload a CSV. Dataloom tests the connection before saving, then reads the schema — tables, views, columns, types and keys — so it knows what is available.
- 2
Pick what to expose
Choose the table or view an endpoint reads from, and which columns it returns. Anything you do not select is unreachable through the API, so exposing one reporting view without opening the rest of the database is straightforward.
- 3
Configure behaviour
Turn on pagination, filtering and sorting, set page sizes and a request timeout, and add a description. Descriptions matter more than they look — they become the documentation and the AI tool description.
- 4
Publish
The endpoint goes live at a predictable URL, the documentation updates itself, and the endpoint becomes available as an MCP tool. Unpublish and it disappears from both just as quickly.
GET /my-api/customers
?filter={"Status":"active","Revenue":{"gte":10000}}
&sort=Revenue&order=desc
&page=1&pageSize=25
X-API-Key: your_api_keyWhat every API comes with
These are not add-ons. Publish an API and all of it exists immediately, on every plan.
Interactive documentation
Browsable API docs served at /{your-api}/docs, generated from your endpoints and always current because they are derived from the same definitions.
OpenAPI description
A standards-compliant OpenAPI JSON document at /{your-api}/openapi.json, ready for code generation or any tool that speaks OpenAPI.
Postman Collection
A downloadable collection with every endpoint, its parameters and pagination options already set up, so testing takes seconds.
API keys
Issue and revoke keys, scope them to an environment, and get reminders before they expire.
Logs and analytics
Request logs and per-endpoint analytics, so you can see what is being called, how often, and how fast.
Rate limiting
Per-endpoint rate limits applied automatically, with the ceiling set by your plan.
When you need more control
Generated endpoints cover most of what an internal API needs, but not everything. For the rest there are custom query endpoints, available on the Starter plan and above, where you write the SQL yourself and bind values with named parameters. Aggregations, multi-table joins and window functions all become endpoints without giving up parameter binding.
SELECT Province,
COUNT(*) AS Customers,
SUM(Revenue12m) AS Revenue
FROM CustomerActivity
WHERE Revenue12m >= :minRevenue
GROUP BY Province
ORDER BY Revenue DESCEnvironments let one API point at different data sources — staging and production, say — with per-datasource overrides and API keys scoped to a single environment. And on Starter and above you can serve an API from your own custom domain, so consumers see your hostname rather than ours.
Every endpoint is also an AI tool
This is the part that tends to surprise people. Publishing an endpoint does not only create a REST route — it also adds a tool to your workspace's hosted MCP server. Claude, ChatGPT, Cursor and any other MCP-compatible client can then call it directly, with typed parameters taken from your real schema.
There is no second integration to build and no separate schema to keep in step. The API you built for your applications is the same API your AI assistant uses, with the same auth rules and the same limits — and it is included on the free plan.
What it costs
The free plan is genuinely usable: one API, one data source, three endpoints and a hosted MCP server, with no card required. Starter is $10 a month and adds write endpoints, custom SQL queries and custom domains. Pro is $25 a month for larger limits and higher rate limits, and Enterprise adds SSO/SAML and an SLA.
Paid plans carry a 14-day money-back guarantee.
Questions, answered
Do I need to write any code?
No. You connect a data source, choose the tables or views to expose, configure each endpoint in the browser and publish. If you do want to write SQL, custom query endpoints are available on the Starter plan and above.
Which data sources can I connect?
PostgreSQL, MySQL, SQL Server, MongoDB, Google Sheets, Excel Online and CSV upload — seven sources in total, all through the same builder.
Does Dataloom copy my data?
No. For databases and spreadsheets, Dataloom connects with the credentials you supply and queries the source directly when a request arrives. CSV is the exception, since an uploaded file is a snapshot by nature.
Do I get API documentation?
Yes, automatically. Every API gets interactive documentation, an OpenAPI JSON description, and a downloadable Postman Collection with endpoints and parameters already configured.
Can the API write data, not just read it?
Yes. Alongside GET endpoints you can publish POST, PATCH and DELETE endpoints on the Starter plan and above, bounded by what the credentials you connected with are permitted to do.
Is there a free plan?
Yes. The free plan includes one API, one data source, three endpoints and a hosted MCP server, with no card required.