KO | EN

GitLite

Minimalist Open Source Repository Search Engine

gitlite — search

#dsh #ai-agents #dsh-plugin #codex #developer-tools #claude-code #windows #desktop-app #typescript #rust #ai-agent #cli

context-builder

★ 29 Open GitHub ↗

Infrastructure Context for your Coding Agents (Claude Code / Cursor)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# droidctx

Infrastructure context builder for Claude Code and coding agents.

Connect your production tools (Grafana, Datadog, Kubernetes, CloudWatch, databases, etc.), extract metadata, and generate structured `.md` files that give coding agents instant context about your infrastructure.

<a href="https://discord.com/invite/AQ3tusPtZn">Join our Discord community</a>

## Quick Start

```bash
# Install (creates an isolated venv at ~/.droidctx automatically)
curl -fsSL https://raw.githubusercontent.com/DrDroidLab/context-builder/main/install.sh | bash

# Or via pipx
pipx install git+https://github.com/DrDroidLab/context-builder.git

# Or manually with a venv
python3 -m venv ~/.droidctx && ~/.droidctx/bin/pip install git+https://github.com/DrDroidLab/context-builder.git && mkdir -p ~/.local/bin && ln -sf ~/.droidctx/bin/droidctx ~/.local/bin/droidctx
```

```bash
# 1. Initialize project (creates ./droidctx-context/)
droidctx init

# 2. Auto-detect credentials from local CLI tools (kubectl, aws, gcloud, az)
droidctx detect

# 3. Add any additional credentials manually
vim ./droidctx-context/credentials.yaml

# 4. Sync infrastructure metadata
droidctx sync

# 5. Add the suggested prompt to your CLAUDE.md
```

## Output Structure

After running `droidctx sync`, your context directory will contain:

```
my-infra/
  resources/
    overview.md                    # Summary of all connected tools
    connectors/
      grafana_prod/
        _summary.md                # Resource counts for this connector
        datasources.md             # Grafana datasources
        dashboards.md              # Dashboard index
        dashboards/
          api-gateway.md           # Individual dashboard with panels & queries
          payment-service.md
        alerts.md                  # Alert rules
      k8s_production/
        _summary.md
        namespaces.md
        deployments.md
        services.md
        ingresses.md
        ...
      datadog_prod/
        _summary.md
        monitors.md
        services.md
        dashboards.md
      postgres_main/
        _summary.md
        tables.md
    cross_references/
      services.md                  # Services seen across multiple connectors
```

## Using with Claude Code

After syncing, add this to your `CLAUDE.md`:

```
My production infrastructure context is in ./my-infra/resources/.
Refer to this when investigating issues, writing queries, or understanding system topology.
```

## Commands

### `droidctx init`

Creates folder structure and a credentials template.

```bash
droidctx init                      # Creates ./droidctx-context/
droidctx init --path ./my-infra    # Custom path
```

### `droidctx detect`

Auto-detects credentials from locally configured CLI tools and populates `credentials.yaml`. Scans for `kubectl`, `aws`, `gcloud`, and `az`, extracts their active configurations, and merges discovered connectors into your credentials file without overwriting existing entries.

```bash
droidctx detect                                # Uses ./droidctx-context/credentials.yaml
droidctx detect --keyfile ./my-infra/creds.yaml # Custom keyfile
```

**What gets detected:**

| CLI Tool | Connectors Created | Values Extracted |
|----------|-------------------|-----------------|
| `kubectl` | KUBERNETES (cli mode) | Cluster name from current context |
| `aws` | CLOUDWATCH, EKS | Region, EKS cluster names |
| `gcloud` | GKE, GCM | Project ID, zone, GKE cluster names |
| `az` | AZURE | Tenant ID, subscription ID (client ID/secret need manual entry) |

Kubernetes connectors created by `detect` use **CLI mode** (`_cli_mode: true`), which means they extract resources directly via `kubectl` using your current kubeconfig context — no API server URL or token needed.

### `droidctx sync`

Connects to your tools, extracts metadata, and generates `.md` context files.

```bash
droidctx sync                                  # Uses ./droidctx-context/credentials.yaml
droidctx sync --keyfile ./my-infra/creds.yaml   # Custom keyfile
droidctx sync --connectors grafana_prod,k8s_prod  # Sync specific connectors
droidctx sync --dry-run                         # Preview what would be synced
droidctx sync --verbose                         # Verbose logging
```

### `droidctx check`

Validates credentials format and checks for required CLI tools.

```bash
droidctx check                                 # Uses ./droidctx-context/credentials.yaml
droidctx check --keyfile ./my-infra/creds.yaml  # Custom keyfile
```

### `droidctx list-connectors`

Shows all supported connector types and their required fields.

```bash
droidctx list-connectors
droidctx list-connectors --type GRAFANA
```

### `droidctx auto-sync`

Keep your context files fresh automatically. Uses launchd on macOS and cron on Linux.

```bash
# Enable auto-sync (runs every 30 minutes by default)
droidctx auto-sync enable --keyfile ./droidctx-context/credentials.yaml

# Custom interval (in minutes)
droidctx auto-sync enable --keyfile ./droidctx-context/credentials.yaml --interval 60

# Check status
droidctx auto-sync status

# Disable
droidctx auto-sync disable
```

Logs are written to `~/.config/droidctx/auto-sync.log`.

> **macOS note:** You may need to allow droidctx in System Settings → Privacy & Security the first time.

## Credentials Format

Create a YAML file with your connector credentials. Run `droidctx init` to generate a template with all supported types, or `droidctx detect` to auto-populate from your CLI tools.

```yaml
# Auto-detected by `droidctx detect` (uses kubectl directly, no token needed)
k8s_my-cluster:
  type: "KUBERNETES"
  _cli_mode: true
  cluster_name: my-cluster

# Auto-detected by `droidctx detect`
cloudwatch_us-east-1:
  type: "CLOUDWATCH"
  region: us-east-1

# Manual entry
grafana_prod:
  type: "GRAFANA"
  grafana_host: https://your-grafana.com
  grafana_api_key: glsa_xxxxxxxxxxxx

datadog_prod:
  type: "DATADOG"
  dd_api_key: your_api_key
  dd_app_key: your_app_key

# Standard Kubernetes (API server + token, without CLI mode)
k8s_production:
  type: "KUBERNETES"
  cluster_name: prod-cluster
  cluster_api_server: https://k8s-api.example.com
  cluster_token: eyJhbGciOiJSUzI1NiIs...

postgres_main:
  type: "POSTGRES"
  host: db.example.com
  port: 5432
  database: production
  user: readonly_user
  password: secret
```

### CLI Mode for Kubernetes

When `_cli_mode: true` is set on a KUBERNETES connector, droidctx uses `kubectl` directly with your current kubeconfig context instead of requiring an API server URL and token. This is the default when connectors are created via `droidctx detect`. Resources extracted: Namespaces, Services, Deployments, Ingresses, StatefulSets, ReplicaSets, HPAs, and NetworkPolicies.

## Supported Connectors (25)

| Category | Connectors |
|----------|-----------|
| **Monitoring** | Grafana, Datadog, New Relic, CloudWatch, SigNoz, Sentry |
| **Kubernetes** | Kubernetes, EKS, GKE |
| **Cloud** | Azure, GCM (Google Cloud Monitoring) |
| **Databases** | PostgreSQL, MongoDB, ClickHouse, Generic SQL |
| **Search** | Elasticsearch, OpenSearch |
| **CI/CD** | GitHub, ArgoCD, Jenkins |
| **Project Management** | Jira Cloud |
| **Logs** | Grafana Loki, Victoria Logs, Coralogix, PostHog |

## Why

When your coding agent debugs production issues, it wastes tokens fumbling across tools, picks wrong MCP servers, and hallucinates about your setup. Pre-built context files fix that — fewer steps, better hypotheses, less noise.

Your agent will know:
- Which dashboards exist and what metrics they track
- What services are running and where they're deployed
- Which alerts are configured and what they monitor
- What database tables/schemas exist
- How to write queries for your specific tools

## Requirements

- Python >= 3.9
- Some connectors require CLI tools: `kubectl` (Kubernetes), `aws` (CloudWatch/EKS), `az` (Azure), `gcloud` (GKE/GCM)

## Community

Join our [Discord community](https://discord.com/invite/AQ3tusPtZn) for support, feedback, and discussions.

## License

MIT

context-builder

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README