# deployment-intelligence-mcp

Turn DevOps configuration into Deployment Knowledge.

## Why does this exist?

CodeGraph answers:

> Who calls `CreateAccountUseCase()`?

Deployment Intelligence MCP answers:

> How does `CreateAccountUseCase` reach Production?

This project exists to connect code and deployment artifacts across the delivery chain.
It is about the knowledge between code, build, workflow, deployment, service, and ingress.
YAML is an implementation detail; the product is Deployment Knowledge.

## What it does today

This repository extracts deployment knowledge from a local repository and caches it in SQLite.
The current implementation supports:

- GitHub Actions workflows under `.github/workflows/`
- Kubernetes manifests under `k8s/`
- a root `Dockerfile`

It exposes a query layer for MCP agents to inspect the knowledge cache,
not to directly search raw YAML.

## High-level flow

```text
Repository
     │
     ▼
Indexer
     │
     ▼
Knowledge DB
     │
     ▼
Query Engine
     │
     ▼
MCP
```

## Primary abstractions

- `Workflow`
- `Job`
- `Step`
- `Image`
- `Deployment`
- `Service`

## Current CLI commands

Install the project:

```bash
python -m pip install -e .[dev]
```

Show CLI help:

```bash
dimcp --help
```

Index the current repository:

```bash
dimcp index .
```

Run the MCP server:

```bash
dimcp serve
```

## Current MCP tools

The server exposes the following tools:

- `list_workflows`
- `get_workflow`
- `list_services`
- `list_deployments`
- `list_images`

## What’s next

Planned v0.1 improvements:

- `dimcp inspect` for quick knowledge summaries
- `dimcp explain <target>` for deployment explanations
- trace-first queries like `trace_service()` and `trace_image()`
- a `.dimcp/` cache location with `graph.db` and metadata
- a stronger bridge from methods to services and back

## Testing

Run the test suite:

```bash
pytest -q
```