# Contextloop [Health: Active]

**Category:** 📊 Data Platforms  
**Repository:** https://getcontextloop.com/  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/contextloop

## Description
Hosted remote MCP server that gives Claude Code, Cursor and other data agents your team's metric definitions, business terms and SQL rules, turns user corrections into reviewable rules, and tracks answer accuracy over time. Exports to dbt MetricFlow and Cube YAML. No warehouse access needed. Try the demo: https://getcontextloop.com/mcp?key=demo

## Tools
Capabilities this server exposes over MCP:

- **get_context** — Call this BEFORE writing SQL about company data. Returns the team's approved metric definitions, business terms, table notes, rules and verified example queries relevant to the question, plus a log_id to use with report_outcome.
- **report_outcome** — Call after the user confirms or corrects an answer. A correction becomes a proposed definition the team can approve, so future answers get it right.
- **propose_definition** — Propose a metric, business term, table note, rule or verified query for the team's semantic layer. Proposals are reviewed by a human before they're served.
- **list_definitions** — List every approved metric, term, table note and rule in the workspace (names only unless full=true).

## Claude Desktop Quick Installation
Remote MCP endpoint (confidence: high). Install path detected from listing signals. Add as a URL/SSE server in your client:

```json
"mcpServers": {
  "contextloop": {
    "url": "https://getcontextloop.com/mcp?key=demo"
  }
}
```

## Documentation

## What Contextloop does

Contextloop provides a hosted remote MCP endpoint for sharing a team’s data context with AI agents. Its focus is the meaning and usage of data: metric definitions, business terms, and SQL rules. This gives agents a reference for interpreting requests and producing answers that follow the team’s established conventions.

The service also turns corrections from users into rules that can be reviewed. Those rules provide a way to capture recurring feedback instead of leaving corrections as isolated conversations. Contextloop tracks answer accuracy over time, allowing teams to monitor whether agent responses are improving against their expectations.

Contextloop is designed for data-agent workflows where definitions and query guidance matter, but direct warehouse connectivity is not required.

## How it works

Agents connect to Contextloop through its hosted remote MCP endpoint. The service supplies the team’s metric, terminology, and SQL guidance to supported agents, which can use that context when answering data-related questions.

When a user corrects an answer, Contextloop records the correction as a reviewable rule. Accuracy tracking then provides an ongoing view of answer quality. Definitions and rules can also be exported in YAML formats used by dbt MetricFlow and Cube, helping carry the same data semantics into those systems.

A demo endpoint is available at `https://getcontextloop.com/mcp?key=demo` for trying the service.

## Setup and configuration

Contextloop is hosted, so it is used as a remote MCP service rather than installed alongside a local warehouse. Connect an MCP-compatible client to the Contextloop endpoint. Claude Code and Cursor are named supported clients, along with other data agents that can connect to a remote MCP server.

The demo URL includes a `demo` key and can be used to try the service. Teams evaluating the service should use the connection details supplied by Contextloop for their own environment rather than relying on the demo endpoint for production workflows.

No warehouse access is needed for Contextloop to provide metric definitions, business terms, SQL rules, correction handling, or accuracy tracking.

## Tools and capabilities

- Provide metric definitions to data agents.
- Supply business terms and SQL rules as agent context.
- Convert user corrections into reviewable rules.
- Track answer accuracy over time.
- Export definitions and rules to dbt MetricFlow YAML.
- Export definitions and rules to Cube YAML.

## Limitations and notes

Contextloop focuses on data context and agent answer quality rather than warehouse access. Its documented integrations for exported definitions are dbt MetricFlow and Cube, using YAML output. The service is hosted remotely, and the available demo connection is intended for evaluation.

_Full upstream README: https://allmcps.com/mcp/contextloop/readme_