Skyline MCP

---

What is Skyline?

Skyline is an MCP (Model Context Protocol) server that turns external APIs into tools AI agents can use. It reads API specifications, normalizes them into a canonical model, and exposes every operation as a callable MCP tool with full JSON Schema validation.

You describe your APIs in a YAML file. Skyline does the rest:

  ┌──────────────┐      ┌────────────────┐      ┌──────────────────┐
  │  API Specs   │ ---> │    Skyline      │ ---> │    MCP Tools     │
  │              │      │                 │      │                  │
  │  OpenAPI     │      │  Auto-detect    │      │  tools/list      │
  │  Swagger 2   │      │  Parse & norm   │      │  tools/call      │
  │  GraphQL     │      │  Validate       │      │  resources/list  │
  │  SOAP/WSDL   │      │  Execute        │      │  resources/read  │
  │  OData v4    │      │                 │      │                  │
  │  gRPC        │      │  stdio / HTTP   │      │  Claude, Cursor  │
  │  JSON-RPC    │      │                 │      │  or any MCP host │
  │  Postman     │      │                 │      │                  │
  │  AsyncAPI    │      │                 │      │                  │
  │  RAML        │      │                 │      │                  │
  │  API Bluepr. │      │                 │      │                  │
  │  Insomnia    │      │                 │      │                  │
  │  Jenkins     │      │                 │      │                  │
  │  Jira Cloud  │      │                 │      │                  │
  │  Google API  │      │                 │      │                  │
  │  CKAN        │      │                 │      │                  │
  └──────────────┘      └────────────────┘      └──────────────────┘

---

🚀 Code Execution (Default)

Skyline uses code execution by default, providing up to 98% cost reduction compared to traditional MCP:

Traditional MCP:

• AI evaluates 100+ tool definitions (~20K tokens)
• Makes 5-7 separate tool calls
• Sends intermediate data back and forth (~5K tokens per call)
• Cost: ~$0.081 per request (Claude Opus)

Code Execution (Default):

• AI receives tool hints (~2K tokens)
• Writes one JavaScript script
• Skyline executes code in embedded Goja JS runtime — zero external dependencies
• Code calls tools internally, filters data locally
• Cost: ~$0.002 per request (97.7% cheaper!)

Monthly savings: $7,900 for 100K requests

Requirements

Code execution uses the embedded Goja JS runtime (a pure-Go JavaScript engine). There are no external dependencies — everything is compiled into the Skyline binary.

curl -fsSL https://skyline.projex.cc/install | bash

Disable Code Execution

# config.yaml
enable_code_execution: false  # Use traditional MCP tools

See the Skyline documentation for full details on code execution.

---

Supported API Types

Skyline auto-detects the spec format. No manual configuration needed.

Protocol	Detection	Notes
OpenAPI 3.x	openapi field in JSON/YAML	Full path, query, header, and body parameter support
Swagger 2.0	swagger field	Automatically converted to OpenAPI 3 internally
GraphQL	SDL files or introspection	Builds typed queries with variable support and selection sets
WSDL 1.1 / SOAP	XML with <definitions>	Generates SOAP envelopes, parses XML responses to JSON
OData v4	CSDL $metadata XML	Generates CRUD operations per EntitySet with OData query options
gRPC	spec_type: grpc in config	Discovers services via gRPC reflection; builds dynamic protobuf messages
OpenRPC / JSON-RPC	openrpc field in JSON	Wraps calls in JSON-RPC 2.0 envelopes; supports rpc.discover
Postman Collections	schema.getpostman.com in JSON	Walks v2.x collection items; supports folders, path/query/header params, body modes
Google API Discovery	discoveryVersion field	Maps Google's discovery format to REST operations
Jenkins 2.545 ⚠️	/api/json object graph	34 operations - Custom implementation. Jobs, builds, pipelines, Blue Ocean, nodes, credentials, plugins, queue. Full CSRF support. See special cases
Slack Web API ⚠️	{"ok":...} response format	23 operations - Custom implementation. Chat, conversations, users, files, reactions, pins, reminders. See special cases
Jira Cloud	*.atlassian.net host	Auto-fetches the official Atlassian OpenAPI spec
AsyncAPI	asyncapi field in JSON/YAML	Event-driven APIs; maps channels and operations to MCP tools
RAML	#%RAML header	RESTful API Modeling Language; full resource/method support
API Blueprint	FORMAT: 1A header	Markdown-based API description; parses resource groups and actions
Insomnia	_type: export in JSON	Insomnia export collections; walks request items with full param support
CKAN Open Data ⚠️	/api/3/action/ endpoint or spec_type: ckan	7 operations — Custom implementation. Dataset search, resource access, datastore queries, organization/tag listing. Compatible with any CKAN 2.x/3.x portal worldwide.

---

🎯 Recommended: Web UI Configuration

The Web UI is the easiest and most secure way to configure Skyline. All configurations are encrypted automatically with AES-256-GCM.

Start the Web UI

# Default mode: HTTP + Admin UI
skyline

# Or with explicit flags
skyline --transport http --admin --bind localhost:8191

# Open browser
# https://localhost:8191/ui/
# https://localhost:8191/admin/

The encryption key is automatically generated on first run and saved to ~/.skyline/skyline.env.

Features

• ✅ Point-and-click configuration - No YAML editing required
• ✅ Automatic encryption - All profiles encrypted at rest with AES-256-GCM
• ✅ API testing - Test endpoints before saving
• ✅ Syntax validation - Catch errors before they break
• ✅ Secure credential storage - Never store plaintext secrets
• ✅ Multiple profiles - Separate configs for dev/staging/prod

Understanding Encryption Key vs Profile Tokens

🔑 Encryption Key (SKYLINE_PROFILES_KEY):

• Encrypts the ENTIRE profiles.enc.yaml file containing ALL your profiles
• Share with your team via secure channels (1Password, Vault, etc.)
• Protects data at rest on disk

🎫 Profile Tokens (per profile):

• Control WHO can access WHICH specific profile
• Give each user only the tokens for profiles they need
• Provide access control and authentication

Team Example:

# Everyone shares the encryption key to decrypt the file
export SKYLINE_PROFILES_KEY=abc123...

# But each person gets different profile tokens:
# - Developer: dev-token-abc (can only access dev-api)
# - DevOps: prod-token-xyz (can only access prod-api)

Security

Profiles are encrypted using:

• Algorithm: AES-256-GCM (Galois/Counter Mode)
• Key size: 256 bits (32 bytes)
• Authentication: Built-in MAC prevents tampering
• Storage: profiles.enc.yaml (encrypted JSON envelope)

See the Skyline documentation for complete configuration documentation.

---

🔧 Alternative: Manual Configuration

For users who prefer command-line tools:

1. Create a config

Config files support both YAML and JSON formats (auto-detected):

YAML (recommended for readability):

# config.yaml
apis:
  - name: petstore
    spec_url: https://petstore3.swagger.io/api/v3/openapi.json
    auth:
      type: api-key
      header: X-API-Key
      value: ${PETSTORE_API_KEY}

JSON (for programmatic generation):

{
  "apis": [
    {
      "name": "petstore",
      "spec_url": "https://petstore3.swagger.io/api/v3/openapi.json",
      "auth": {
        "type": "api-key",
        "header": "X-API-Key",
        "value": "${PETSTORE_API_KEY}"
      }
    }
  ]
}

Secrets use ${ENV_VAR} syntax and are automatically redacted from all logs.

2. Run

# stdio (for Claude Desktop, Cursor, etc.)
go run ./cmd/skyline --config ./config.yaml

# streamable HTTP (for networked MCP clients)
go run ./cmd/skyline --config ./config.yaml --transport http --bind localhost:8191

3. Connect your AI

Add Skyline to your MCP client config. For Claude Desktop:

{
  "mcpServers": {
    "skyline": {
      "command": "./bin/skyline",
      "args": ["--config", "./config.yaml"]
    }
  }
}

That's it. Your AI agent now has typed, validated tools for every API endpoint.

---

Architecture

Skyline is built around three components:

Skyline MCP Server  cmd/skyline

The core. Loads your config, fetches and parses API specs, builds MCP tools, and serves them over stdio or HTTP.

Config (YAML)
  → Spec Fetcher (URL or file)
    → Auto-Detect adapter (OpenAPI | Swagger2 | GraphQL | WSDL | OData | OpenRPC | Postman | gRPC | AsyncAPI | RAML | API Blueprint | Insomnia | Jenkins | Google | Jira)
      → Canonical Model (Service → Operations → Parameters + Schemas)
        → MCP Registry (tools + resources + JSON Schema validators)
          → MCP Server (JSON-RPC 2.0 over stdio or streamable HTTP)
            → Runtime Executor (HTTP requests, gRPC calls, JSON-RPC envelopes, auth, retries)

Admin UI & Profile Management

Built-in web interface for managing configurations, profiles, and server settings. Accessible via --admin flag (enabled by default).

# Start with admin UI (default)
skyline

# Or explicitly
skyline --transport http --admin --bind localhost:8191

# Open https://localhost:8191/ui/
# Open https://localhost:8191/admin/

Features:

• Profile management — Create named profiles (dev, staging, prod) with encrypted storage
• AES-GCM encryption — All profiles encrypted at rest, key auto-generated
• Spec detection — Auto-discover API specs from base URLs
• Settings editor — Edit server config.yaml via Web UI
• Metrics & audit — View API call history and performance stats

---

Configuration Reference

Full config example

apis:
  - name: petstore-openapi
    spec_url: http://localhost:9999/openapi/openapi.json
    base_url_override: http://localhost:9999/openapi
    auth:
      type: bearer
      token: ${PETSTORE_TOKEN}

  - name: plants-soap
    spec_url: http://localhost:9999/wdsl/wsdl
    auth:
      type: bearer
      token: ${PLANTS_TOKEN}

  - name: cars-graphql
    spec_url: http://localhost:9999/graphql/schema
    base_url_override: http://localhost:9999/graphql
    auth:
      type: basic
      username: ${GRAPHQL_USER}
      password: ${GRAPHQL_PASS}

  - name: jira-cloud
    spec_url: https://your-domain.atlassian.net
    base_url_override: https://your-domain.atlassian.net
    auth:
      type: basic
      username: ${JIRA_EMAIL}
      password: ${JIRA_API_TOKEN}

  - name: movies-odata
    spec_url: http://localhost:9999/odata/$metadata
    base_url_override: http://localhost:9999/odata
    auth:
      type: bearer
      token: ${ODATA_TOKEN}

  - name: calculator-jsonrpc
    spec_url: http://localhost:9999/jsonrpc/openrpc.json
    base_url_override: http://localhost:9999/jsonrpc
    auth:
      type: api-key
      header: X-API-Key
      value: ${JSONRPC_KEY}

  - name: clothes-grpc
    spec_type: grpc
    base_url_override: localhost:50051
    auth:
      type: bearer
      token: ${GRPC_TOKEN}

  - name: jenkins
    spec_url: https://jenkins.example.com/api/json
    base_url_override: https://jenkins.example.com
    auth:
      type: basic
      username: ${JENKINS_USER}
      password: ${JENKINS_API_TOKEN}  # Generate at /me/configure

timeout_seconds: 10
retries: 1

Auth types

Type	Fields
bearer	token
basic	username, password
api-key	header, value

API config fields

Field	Required	Description
name	yes	Unique name for this API (used as tool name prefix)
spec_url	yes*	URL or file path to the API spec
spec_type	no	Force spec type instead of auto-detect. Currently only grpc is supported
base_url_override	no*	Override the base URL from the spec. Required for gRPC (host:port)
auth	no	Authentication config (see auth types below)
jenkins	no	Jenkins-specific config for write operations

* spec_url is not required when spec_type: grpc is set (uses live reflection instead).

MCP server flags

Flag	Default	Description
--config	./config.yaml	Path to config file (YAML or JSON, auto-detected)
--transport	http	stdio or http
--bind	localhost:8191	Listen address for HTTP transport
--admin	true	Enable Web UI and admin dashboard (HTTP only)

Additional flags

Flag	Default	Description
--storage	./profiles.enc.yaml	Encrypted storage path
--auth-mode	bearer	bearer or none
--key-env	SKYLINE_PROFILES_KEY	Env var holding the 32-byte AES key
--env-file		Optional .env file to load

---

Transport Modes

Skyline supports multiple transport modes:

# HTTP + Admin UI (default)
skyline
# or explicitly:
skyline --transport http --admin

# HTTP only (no UI)
skyline --transport http --admin=false

# STDIO for Claude Desktop (coming soon)
skyline --transport stdio

---

Project Layout

skyline-mcp/
│
├── cmd/                              # ── Entrypoint ─────────────────
│   └── skyline/                      #    Unified binary
│       ├── server.go                 #      All transports (http, stdio)
│       └── ui/                       #      Embedded Web UI & admin
│           ├── index.html
│           ├── admin.html
│           ├── app.js
│           └── styles.css
│
├── internal/                         # ── Core ───────────────────────
│   ├── canonical/                    #    Unified API model
│   │   ├── types.go                  #      Service, Operation, Parameter
│   │   └── naming.go                 #      Tool name generation
│   ├── config/                       #    Configuration
│   │   ├── config.go                 #      Types, validation, defaults
│   │   ├── load.go                   #      YAML file loading
│   │   ├── parse.go                  #      YAML parsing
│   │   ├── env.go                    #      ${ENV_VAR} expansion
│   │   └── remote.go                 #      Config server profile fetching
│   ├── mcp/                          #    MCP Protocol
│   │   ├── server.go                 #      JSON-RPC 2.0 handler (stdio)
│   │   ├── http_sse.go               #      Streamable HTTP + SSE transport
│   │   └── registry.go               #      Tool & resource registry
│   ├── runtime/                      #    Execution
│   │   └── executor.go               #      HTTP client, auth, retries
│   ├── redact/                       #    Security
│   │   └── redact.go                 #      Secret redaction for logs
│   │
│   ├── spec/                         # ── Spec Pipeline ──────────────
│   │   ├── adapter.go                #      SpecAdapter interface
│   │   ├── loader.go                 #      Auto-detect + parse orchestrator
│   │   ├── fetch.go                  #      Spec URL fetcher
│   │   ├── openapi_adapter.go        #      OpenAPI 3.x adapter
│   │   ├── swagger2_adapter.go       #      Swagger 2.0 adapter
│   │   ├── graphql_adapter.go        #      GraphQL adapter
│   │   ├── graphql_introspection.go  #      GraphQL introspection query
│   │   ├── wsdl_adapter.go           #      WSDL / SOAP adapter
│   │   ├── odata_adapter.go          #      OData v4 adapter
│   │   ├── openrpc_adapter.go        #      OpenRPC / JSON-RPC adapter
│   │   ├── postman_adapter.go        #      Postman Collections adapter
│   │   ├── grpc_adapter.go           #      gRPC adapter
│   │   ├── google_adapter.go         #      Google API Discovery adapter
│   │   ├── jenkins_adapter.go        #      Jenkins adapter
│   │   ├── jenkins_writes.go         #      Jenkins write operations
│   │   ├── asyncapi_adapter.go       #      AsyncAPI adapter
│   │   ├── raml_adapter.go           #      RAML adapter
│   │   ├── apiblueprint_adapter.go   #      API Blueprint adapter
│   │   └── insomnia_adapter.go       #      Insomnia adapter
│   │
│   └── parsers/                      # ── Parsers ────────────────────
│       ├── openapi/                  #      OpenAPI 3.x parser
│       ├── swagger2/                 #      Swagger 2.0 parser
│       ├── graphql/                  #      GraphQL SDL + introspection
│       ├── wsdl/                     #      WSDL 1.1 parser
│       ├── odata/                    #      OData v4 CSDL parser
│       ├── openrpc/                  #      OpenRPC / JSON-RPC parser
│       ├── postman/                  #      Postman Collection v2.x parser
│       ├── grpc/                     #      gRPC reflection parser
│       ├── googleapi/                #      Google API Discovery parser
│       ├── jenkins/                  #      Jenkins object graph parser
│       ├── asyncapi/                 #      AsyncAPI parser
│       ├── raml/                     #      RAML parser
│       ├── apiblueprint/             #      API Blueprint parser
│       └── insomnia/                 #      Insomnia collection parser
│
├── examples/                         # ── Examples ───────────────────
│   ├── config.yaml.example           #    Full config with all API types
│   ├── config.mock.yaml              #    Config for mock-api server
│   ├── config.yaml                   #    Minimal working config
│   └── skyline.env.example           #    Server env template
│
├── assets/                           # ── Branding ───────────────────
│   ├── skyline-banner.svg
│   └── skyline-logo.svg
│
├── Dockerfile                        # ── Docker ──────────────────────
├── go.mod
├── go.sum
└── README.md

---

Jenkins 2.545 Integration

Skyline provides comprehensive support for Jenkins 2.x with 34 operations covering all major Jenkins APIs. Perfect for CI/CD automation, build management, and pipeline orchestration.

Quick Start

apis:
  - name: jenkins
    spec_url: https://jenkins.example.com/api/json
    base_url_override: https://jenkins.example.com
    auth:
      type: basic
      username: admin
      password: ${JENKINS_API_TOKEN}

Features

• ✅ Auto-detects Jenkins 2.x - No manual configuration
• ✅ 34 operations across 10 API categories
• ✅ Automatic CSRF handling - Crumb fetching and caching
• ✅ Parameterized builds - Full parameter support
• ✅ Pipeline support - Jenkinsfile creation, replay, stages
• ✅ Blue Ocean API - Modern Pipeline visualization

Operations Coverage

Category	Operations	Examples
Core (3)	Root, object queries, version	jenkins__getVersion, jenkins__getRoot
Job Management (9)	CRUD, enable/disable, copy	jenkins__createJob, jenkins__listJobs
Build Operations (7)	Trigger, stop, logs, artifacts	jenkins__triggerBuild, jenkins__getBuildLog
Pipeline (3)	Create, replay, stages	jenkins__createPipeline, jenkins__getPipelineStages
Queue (2)	View, cancel	jenkins__getQueue, jenkins__cancelQueueItem
Nodes (4)	List, configure, offline, delete	jenkins__listNodes, jenkins__markNodeOffline
Credentials (1)	List stores	jenkins__listCredentials
Plugins (1)	List installed	jenkins__listPlugins
Blue Ocean (2)	Pipelines, runs	jenkins__blueOceanPipelines
Users (2)	User info	jenkins__getCurrentUser, jenkins__getUser

Example Usage

List all jobs:

{
  "tool": "jenkins__listJobs",
  "arguments": {}
}

Trigger a parameterized build:

{
  "tool": "jenkins__triggerBuildWithParameters",
  "arguments": {
    "jobName": "deploy-prod",
    "parameters": {
      "BRANCH": "main",
      "ENVIRONMENT": "production"
    }
  }
}

Get build console log:

{
  "tool": "jenkins__getBuildLog",
  "arguments": {
    "jobName": "my-job",
    "buildNumber": "lastBuild"
  }
}

Deployment Example

Kubernetes (Tested with Jenkins 2.545):

apiVersion: apps/v1
kind: Deployment
metadata:
  name: jenkins
spec:
  template:
    spec:
      containers:
      - name: jenkins
        image: jenkins/jenkins:2.545-jdk21
        ports:
        - containerPort: 8080

See the Skyline documentation for complete Jenkins integration docs including all 34 operations with examples, authentication guide, CSRF handling, and troubleshooting.

---

Special Cases

Some APIs don't provide machine-readable specifications (OpenAPI, GraphQL schema, etc.) or have specification issues that prevent auto-detection. For these, Skyline includes custom adapters that manually define operations based on official API documentation.

Jenkins 2.545 ⚠️

Why custom?
Jenkins doesn't expose an API specification. The /api/json endpoint returns current state (jobs, builds, queues), not endpoint definitions. There's no machine-readable documentation of available operations, parameters, or schemas.

Solution:
Manually implemented 34 operations based on Jenkins REST API documentation, including:

• Automatic CSRF crumb handling
• Parameterized build support
• Pipeline operations (Jenkinsfile, replay, stages)
• Blue Ocean API integration
• Node/agent management

See: Skyline documentation for full Jenkins integration details.

Slack Web API ⚠️

Why custom?
Slack provides an OpenAPI spec (Swagger 2.0) at https://api.slack.com/specs/openapi/v2/slack_web.json, but it contains validation errors that break standard Swagger → OpenAPI 3 conversion:

• Uses non-standard array-based items for union types (oneOf)
• Example: "items": [{"$ref": "..."}, {"type": "null"}] (should be object)
• 3 definitions affected: objs_conversation, objs_response_metadata, objs_user

Solution:
Manually implemented 23 most-used operations based on Slack Web API documentation, including:

Operations (23):

• Chat (4): postMessage, update, delete, getPermalink
• Conversations (6): list, create, history, invite, info, archive
• Users (3): list, info, conversations
• Files (2): upload, list
• Reactions (2): add, remove
• Pins (3): add, remove, list
• Reminders (2): add, list
• Channels (1): list (legacy)

Auth:
Bearer token (bot token or user token with appropriate scopes).

Example:

apis:
  - name: slack
    spec_url: https://slack.com/api/auth.test  # Any Slack endpoint (for detection)
    base_url_override: https://slack.com/api
    auth:
      type: bearer
      token: ${SLACK_BOT_TOKEN}

Future: If Slack fixes their OpenAPI spec validation issues, Skyline can switch to auto-detection for full 172-operation coverage.

CKAN Open Data Portals ⚠️

Why custom? CKAN (https://ckan.org) is the open-source data management system powering the majority of government open data portals worldwide. It does not expose an OpenAPI, Swagger, or any machine-readable specification — it uses its own fixed action-based JSON API at /api/3/action/{action}.

Solution: Manually implemented 7 core operations based on the CKAN API documentation, covering dataset discovery, resource access, and datastore queries.

Operations (7):

• searchDatasets — Full-text search with filter queries, sorting, and pagination
• listDatasets — List all published dataset names with pagination
• getDataset — Full dataset metadata including all resources
• getResource — Individual resource (file/link) metadata and download URL
• queryDatastore — Filtered SQL-style queries against tabular data resources
• listOrganizations — All data publishers/organizations on the portal
• listTags — All subject tags used to categorize datasets

Compatible portals (any CKAN 2.x / 3.x instance):

• 🇺🇸 data.gov — US Federal Open Data
• 🇬🇧 data.gov.uk — UK Government Data
• 🇸🇦 open.data.gov.sa — Saudi Arabia Open Data
• 🇪🇺 data.europa.eu — EU Open Data Portal
• 🇦🇺 data.gov.au — Australian Government Data
• 🇨🇦 open.canada.ca — Canada Open Data
• 🇮🇳 data.gov.in — India Government Data
• 🇧🇷 dados.gov.br — Brazil Open Data
• And 100+ more national, regional, and city portals

Example:

apis:
  - name: us-open-data
    spec_type: ckan
    base_url_override: https://catalog.data.gov
    # No auth needed for public portals
    # auth required only for private datasets:
    # auth:
    #   type: api-key
    #   header: Authorization
    #   value: ${CKAN_API_TOKEN}

Auto-detection: Skyline can also auto-detect CKAN portals — just provide the base URL and Skyline will probe /api/3/action/package_list to confirm it's a CKAN instance.

---

Building

# Build the binary
go build -o ./bin/skyline ./cmd/skyline

# Run tests
go test ./...

# Build from source (install script does this)
curl -fsSL https://skyline.projex.cc/install | bash -s source

Docker

# Build the container
docker build -t skyline-mcp .

# Run with a config file
docker run -p 8191:8191 -v $(pwd)/config.yaml:/app/config.yaml skyline-mcp

---

Troubleshooting

Problem	Fix
Startup fails with "missing env"	Ensure all ${ENV_VAR} references in your config have corresponding environment variables set
Spec load fails	Verify the spec_url is reachable; use the config server's test feature to check
Tool call returns auth errors	Double-check your auth block; bearer tokens, API keys, and basic auth credentials must match what the target API expects
SOAP responses look like raw XML	This is expected for non-standard SOAP services; Skyline parses standard SOAP envelopes automatically

---

Examples & Testing

Mocking Bird - Test API Server

Repository: github.com/emadomedher/mocking-bird

Mocking Bird is a standalone mock API server that implements all 7 protocol types supported by Skyline. It's the perfect testing companion for validating Skyline's functionality without needing real API credentials or external dependencies.

Why Mocking Bird?

1. Protocol Coverage Testing
Test every spec adapter (OpenAPI, GraphQL, WSDL, OData, etc.) locally without external services:

apis:
  - name: pets-openapi
    spec_url: http://localhost:9999/openapi/openapi.json
  - name: cars-graphql
    spec_url: http://localhost:9999/graphql/schema
  - name: plants-soap
    spec_url: http://localhost:9999/wdsl/wsdl
  - name: movies-odata
    spec_url: http://localhost:9999/odata/$metadata
  - name: calculator-jsonrpc
    spec_url: http://localhost:9999/jsonrpc/openrpc.json

2. Integration Testing
Validate Skyline's MCP tool generation, JSON Schema validation, and runtime execution without hitting production APIs.

3. Development Workflow
Develop and test new Skyline features (auth handling, error recovery, spec parsing) against a predictable, controllable API environment.

4. CI/CD Pipeline Integration
Run automated tests in CI without API keys, rate limits, or network flakiness.

Mocking Bird API Coverage

Mock API	Protocol	Port	Spec Endpoint
Pets	OpenAPI 3.x	9999	/openapi/openapi.json
Dinosaurs	Swagger 2.0	9999	/swagger/swagger.json
Plants	WSDL/SOAP	9999	/wdsl/wsdl
Cars	GraphQL	9999	/graphql/schema
Movies	OData v4	9999	/odata/$metadata
Calculator	JSON-RPC 2.0	9999	/jsonrpc/openrpc.json
Clothes	gRPC	50051-54	Server reflection enabled

Quick Start with Mocking Bird

# Clone and run Mocking Bird
git clone https://github.com/emadomedher/mocking-bird
cd mocking-bird
go run .
# Listening on http://localhost:9999

# In another terminal, run Skyline with mock config
cd skyline-mcp
go run ./cmd/skyline --config ./examples/config.mock.yaml

Example Config for Mocking Bird

See examples/config.mock.yaml for a complete working config that uses all Mocking Bird endpoints:

apis:
  - name: pets-openapi
    spec_url: http://localhost:9999/openapi/openapi.json
    base_url_override: http://localhost:9999/openapi

  - name: dinosaurs-swagger
    spec_url: http://localhost:9999/swagger/swagger.json
    base_url_override: http://localhost:9999/swagger
    auth:
      type: bearer
      token: MOCK_DINO_TOKEN

  - name: plants-soap
    spec_url: http://localhost:9999/wdsl/wsdl
    auth:
      type: bearer
      token: MOCK_TOKEN

  - name: cars-graphql
    spec_url: http://localhost:9999/graphql/schema
    base_url_override: http://localhost:9999/graphql
    auth:
      type: basic
      username: graphql-user
      password: MOCK_GRAPHQL_PASS

  - name: movies-odata
    spec_url: http://localhost:9999/odata/$metadata
    base_url_override: http://localhost:9999/odata

  - name: calculator-jsonrpc
    spec_url: http://localhost:9999/jsonrpc/openrpc.json
    base_url_override: http://localhost:9999/jsonrpc
    auth:
      type: api-key
      header: X-API-Key
      value: calc-key

Test All Tools

Once running, use the MCP inspector or any MCP client to verify:

# List all generated tools
mcpx tools list

# Test OpenAPI (Pets)
mcpx tools call pets-openapi__listPets

# Test GraphQL (Cars)
mcpx tools call cars-graphql__query_listCars

# Test SOAP (Plants)
mcpx tools call plants-soap__getAllPlants

# Test OData (Movies)
mcpx tools call movies-odata__listMovies

# Test JSON-RPC (Calculator)
mcpx tools call calculator-jsonrpc__add '{"a": 5, "b": 3}'

Result: Full end-to-end validation of Skyline's spec parsing, tool generation, and runtime execution across all supported protocols.

---

Uninstall

To completely remove Skyline from your system:

curl -fsSL https://skyline.projex.cc/uninstall | bash

This will:

• Stop and disable systemd service
• Remove service file
• Remove binary from /usr/local/bin or ~/.local/bin
• Preserve your data in ~/.skyline/ (profiles, config, audit DB)

Manual removal:

# Stop service
systemctl --user stop skyline
systemctl --user disable skyline

# Remove service file
rm -f ~/.config/systemd/user/skyline.service
systemctl --user daemon-reload

# Remove binary
sudo rm -f /usr/local/bin/skyline
rm -f ~/.local/bin/skyline

# Optional: remove data
rm -rf ~/.skyline/

---

---

License

GNU Affero General Public License v3.0 (AGPL-3.0)

Skyline is licensed under the AGPL-3.0. See LICENSE for the full license text.

Why AGPL-3.0?

• ✅ Strong copyleft: Ensures the software and all modifications remain open source
• ✅ Network protection: Users interacting over a network get access to the source code
• ✅ Commercial use: Can be used commercially, but modifications must be shared
• ✅ Community first: Contributions benefit everyone in the ecosystem

Copyright 2026 Projex Digital