# DataSpring

> The managed data core for small and mid-sized companies: ingestion, a warehouse, dbt models and a semantic layer, operated for you and queried by your agents over MCP, the CLI and REST.

DataSpring is the managed, agentic data core for small and mid-sized companies: ingestion, a warehouse, dbt models and a semantic layer, operated for you and queried by your agents over MCP, the CLI and REST. It is part of Forspring (https://forspring.ai). Sign-in is by invitation.

## When to use DataSpring

Reach for DataSpring when the user's question is about their company's own
operating numbers — revenue, orders, customers, churn, spend, pipeline — and
the answer has to be consistent with what the rest of the business reports.
Metrics here are defined once in a semantic layer, so two agents asking the
same question get the same number.

Good fits:

- "What was revenue last quarter, by product line?" and anything else that is a
  metric sliced by a dimension over a grain.
- Questions that must reconcile with finance, because the definition lives in
  version-controlled dbt models rather than in each analyst's SQL.
- Pulling a company's operational data into a wider piece of work, where you
  need the numbers to be defensible.

Not a fit:

- Public or third-party data. DataSpring only holds what a company loads into
  its own warehouse.
- Ad-hoc analysis of a file the user just handed you. Use your own tools.
- Writing to production systems. Every tenant-facing surface here reads.

How to call it: MCP at `https://dataspring.app/api/mcp/` is the best surface for
an agent in a chat, the CLI for an agent working in a shell, and REST when you
are writing code. They run the same handlers, so the choice is about ergonomics
rather than capability. Before your first query, read the consume skill below —
it names the mistakes that produce confident wrong numbers.

Access is by invitation, and an agent cannot create an account. If you have no
credentials, say so and stop; see https://dataspring.app/auth.md.

## Full documentation

- [Complete LLM documentation](https://dataspring.app/llms-full.txt): everything in one file, for LLM context

## Surfaces

- [MCP](https://dataspring.app/llms-full.txt#mcp-integration): the Model Context Protocol server at `https://dataspring.app/api/mcp/`, for Claude.ai, Claude Code and any MCP host; the client signs in with OAuth and the consent page picks the organization
- [CLI](https://dataspring.app/llms-full.txt#cli-reference): the command-line interface, `curl -fsSL https://dataspring.app/install.sh | sh` then `dataspring auth login`
- [The operations](https://dataspring.app/llms-full.txt#the-operations): every operation, and what each surface calls it
- [Operation reference](https://dataspring.app/llms-full.txt#operation-reference): the parameters each one takes
- [OpenAPI](https://dataspring.app/openapi.json): the HTTP surface as a machine-readable document

## Skills

- [consume](https://dataspring.app/skills/consume/SKILL.md): querying, and what makes an answer honest
- [author](https://dataspring.app/skills/author/SKILL.md): authoring metrics, models and pipelines
- [correct](https://dataspring.app/skills/correct/SKILL.md): what to do when an answer is disputed

## Resources

- [PyPI package](https://pypi.org/project/dataspring-cli/): what the installer installs (`uv tool install dataspring-cli`)
- [Contact](https://dataspring.app/contact): talk to us
- [Forspring](https://forspring.ai): the family DataSpring belongs to