mcp-builder

Name: mcp-builder
Author: Rune-kit

80stars21forks

How to install mcp-builder?

rune-kit/rune/mcp-builder

$curl -o .claude/agents/mcp-builder.md https://raw.githubusercontent.com/Rune-kit/rune/HEAD/agents/mcp-builder.md

Installs into the current project.

Summary

Build MCP servers from specs — generates tool definitions, resource handlers, test suites. Supports TypeScript (official SDK) and Python (FastMCP). Multi-provider adapter pattern.

Files · 1

View on GitHub

agents/mcp-builder.md

1	# mcp-builder
2
3	## Purpose
4
5	MCP server builder. Generates complete, tested MCP servers from a natural language description or specification. Handles tool definitions, resource handlers, input validation, error handling, configuration, tests, and documentation. Supports TypeScript (official SDK) and Python (FastMCP).
6
7	## Triggers
8
9	- Called by `cook` when MCP-related task detected (keywords: "MCP server", "MCP tool", "model context protocol")
10	- Called by `scaffold` when MCP Server template selected
11	- `/rune mcp-builder <description>` — manual invocation
12	- Auto-trigger: when project contains `mcp.json`, `@modelcontextprotocol/sdk`, or `fastmcp` in dependencies
13
14	## Calls (outbound)
15
16	- `ba` (L2): if user description is vague — elicit requirements for what tools/resources the server should expose
17	- `research` (L3): look up target API documentation, existing MCP servers for reference
18	- `test` (L2): generate and run test suite for the server
19	- `docs` (L2): generate server documentation (tool catalog, installation, configuration)
20	- `verification` (L3): verify server builds and tests pass
21
22	## Called By (inbound)
23
24	- `cook` (L1): when MCP-related task detected
25	- `scaffold` (L1): MCP Server template in Phase 5
26	- User: `/rune mcp-builder` direct invocation
27
28	## Executable Steps
29
30	### Step 1 — Spec Elicitation
31
32	If description is detailed enough (tools, resources, target API specified), proceed.
33	If vague, ask targeted questions:
34
35	1. What tools should this MCP server expose? (actions the AI can perform)
36	2. What resources does it manage? (data the AI can read)
37	3. What external APIs does it connect to? (if any)
38	4. TypeScript or Python? (default: TypeScript with @modelcontextprotocol/sdk)
39	5. Authentication? (API keys, OAuth, none)
40
41	If user provides a detailed spec or existing API docs → extract answers, confirm.
42
43	### Step 2 — Architecture Design
44
45	<MUST-READ path="references/auto-discovery-pattern.md" trigger="when the server has 5+ tools OR multiple API providers — use auto-discovery registry for graceful degradation"/>
46
47	Determine server structure based on spec:
48
49	TypeScript (default):
50	```
51	mcp-server-<name>/
52	├── src/
53	│ ├── index.ts — server entry point, tool/resource registration
54	│ ├── tools/
55	│ │ ├── <tool-name>.ts — one file per tool
56	│ │ └── index.ts — tool registry
57	│ ├── resources/
58	│ │ ├── <resource>.ts — one file per resource type
59	│ │ └── index.ts — resource registry
60	│ ├── lib/
61	│ │ ├── client.ts — external API client (if applicable)
62	│ │ └── types.ts — shared types
63	│ └── config.ts — environment variable validation
64	├── tests/
65	│ ├── tools/
66	│ │ └── <tool-name>.test.ts
67	│ └── resources/
68	│ └── <resource>.test.ts
69	├── package.json
70	├── tsconfig.json
71	├── .env.example
72	└── README.md
73	```
74
75	Python (FastMCP):
76	```
77	mcp-server-<name>/
78	├── src/
79	│ ├── server.py — FastMCP server with tool/resource decorators
80	│ ├── tools/
81	│ │ └── <tool_name>.py
82	│ ├── resources/
83	│ │ └── <resource>.py
84	│ ├── lib/
85	│ │ ├── client.py — external API client
86	│ │ └── types.py — Pydantic models
87	│ └── config.py — settings via pydantic-settings
88	├── tests/
89	│ ├── test_<tool_name>.py
90	│ └── test_<resource>.py
91	├── pyproject.toml
92	├── .env.example
93	└── README.md
94	```
95
96	### Step 3 — Generate Server Code
97
98	#### Tool Generation
99
100	For each tool:
101
102	TypeScript:
103	```typescript
104	import { z } from 'zod';
105
106	export const toolName = {
107	name: 'tool_name',
108	description: 'What this tool does — used by AI to decide when to call it',
109	inputSchema: z.object({
110	param1: z.string().describe('Description for AI'),
111	param2: z.number().optional().describe('Optional parameter'),
112	}),
113	async handler(input: { param1: string; param2?: number }) {
114	// Implementation
115	return { content: [{ type: 'text', text: JSON.stringify(result) }] };
116	},
117	};
118	```
119
120	Python (FastMCP):
121	```python
122	from fastmcp import FastMCP
123
124	mcp = FastMCP("server-name")
125
126	@mcp.tool()
127	async def tool_name(param1: str, param2: int \| None = None) -> str:
128	"""What this tool does — used by AI to decide when to call it."""
129	# Implementation
130	return json.dumps(result)
131	```
132
133	#### Resource Generation
134
135	For each resource:
136	- URI template with parameters
137	- Read handler that returns structured content
138	- List handler for collections
139
140	#### Configuration
141
142	Generate `.env.example` with all required environment variables:
143	```env
144	# Required
145	API_KEY=your_api_key_here
146	API_BASE_URL=https://api.example.com
147
148	# Optional
149	LOG_LEVEL=info
150	CACHE_TTL=300
151	```
152
153	Genera

People who liked this also viewed

6 picks

Type

mcp-builder

80stars21forks

View on GitHub

How to install mcp-builder?

rune-kit/rune/mcp-builder

$curl -o .claude/agents/mcp-builder.md https://raw.githubusercontent.com/Rune-kit/rune/HEAD/agents/mcp-builder.md

Installs into the current project.

Summary

Build MCP servers from specs — generates tool definitions, resource handlers, test suites. Supports TypeScript (official SDK) and Python (FastMCP). Multi-provider adapter pattern.

Files · 1

View on GitHub

agents/mcp-builder.md

1	# mcp-builder
2
3	## Purpose
4
5	MCP server builder. Generates complete, tested MCP servers from a natural language description or specification. Handles tool definitions, resource handlers, input validation, error handling, configuration, tests, and documentation. Supports TypeScript (official SDK) and Python (FastMCP).
6
7	## Triggers
8
9	- Called by `cook` when MCP-related task detected (keywords: "MCP server", "MCP tool", "model context protocol")
10	- Called by `scaffold` when MCP Server template selected
11	- `/rune mcp-builder <description>` — manual invocation
12	- Auto-trigger: when project contains `mcp.json`, `@modelcontextprotocol/sdk`, or `fastmcp` in dependencies
13
14	## Calls (outbound)
15
16	- `ba` (L2): if user description is vague — elicit requirements for what tools/resources the server should expose
17	- `research` (L3): look up target API documentation, existing MCP servers for reference
18	- `test` (L2): generate and run test suite for the server
19	- `docs` (L2): generate server documentation (tool catalog, installation, configuration)
20	- `verification` (L3): verify server builds and tests pass
21
22	## Called By (inbound)
23
24	- `cook` (L1): when MCP-related task detected
25	- `scaffold` (L1): MCP Server template in Phase 5
26	- User: `/rune mcp-builder` direct invocation
27
28	## Executable Steps
29
30	### Step 1 — Spec Elicitation
31
32	If description is detailed enough (tools, resources, target API specified), proceed.
33	If vague, ask targeted questions:
34
35	1. What tools should this MCP server expose? (actions the AI can perform)
36	2. What resources does it manage? (data the AI can read)
37	3. What external APIs does it connect to? (if any)
38	4. TypeScript or Python? (default: TypeScript with @modelcontextprotocol/sdk)
39	5. Authentication? (API keys, OAuth, none)
40
41	If user provides a detailed spec or existing API docs → extract answers, confirm.
42
43	### Step 2 — Architecture Design
44
45	<MUST-READ path="references/auto-discovery-pattern.md" trigger="when the server has 5+ tools OR multiple API providers — use auto-discovery registry for graceful degradation"/>
46
47	Determine server structure based on spec:
48
49	TypeScript (default):
50	```
51	mcp-server-<name>/
52	├── src/
53	│ ├── index.ts — server entry point, tool/resource registration
54	│ ├── tools/
55	│ │ ├── <tool-name>.ts — one file per tool
56	│ │ └── index.ts — tool registry
57	│ ├── resources/
58	│ │ ├── <resource>.ts — one file per resource type
59	│ │ └── index.ts — resource registry
60	│ ├── lib/
61	│ │ ├── client.ts — external API client (if applicable)
62	│ │ └── types.ts — shared types
63	│ └── config.ts — environment variable validation
64	├── tests/
65	│ ├── tools/
66	│ │ └── <tool-name>.test.ts
67	│ └── resources/
68	│ └── <resource>.test.ts
69	├── package.json
70	├── tsconfig.json
71	├── .env.example
72	└── README.md
73	```
74
75	Python (FastMCP):
76	```
77	mcp-server-<name>/
78	├── src/
79	│ ├── server.py — FastMCP server with tool/resource decorators
80	│ ├── tools/
81	│ │ └── <tool_name>.py
82	│ ├── resources/
83	│ │ └── <resource>.py
84	│ ├── lib/
85	│ │ ├── client.py — external API client
86	│ │ └── types.py — Pydantic models
87	│ └── config.py — settings via pydantic-settings
88	├── tests/
89	│ ├── test_<tool_name>.py
90	│ └── test_<resource>.py
91	├── pyproject.toml
92	├── .env.example
93	└── README.md
94	```
95
96	### Step 3 — Generate Server Code
97
98	#### Tool Generation
99
100	For each tool:
101
102	TypeScript:
103	```typescript
104	import { z } from 'zod';
105
106	export const toolName = {
107	name: 'tool_name',
108	description: 'What this tool does — used by AI to decide when to call it',
109	inputSchema: z.object({
110	param1: z.string().describe('Description for AI'),
111	param2: z.number().optional().describe('Optional parameter'),
112	}),
113	async handler(input: { param1: string; param2?: number }) {
114	// Implementation
115	return { content: [{ type: 'text', text: JSON.stringify(result) }] };
116	},
117	};
118	```
119
120	Python (FastMCP):
121	```python
122	from fastmcp import FastMCP
123
124	mcp = FastMCP("server-name")
125
126	@mcp.tool()
127	async def tool_name(param1: str, param2: int \| None = None) -> str:
128	"""What this tool does — used by AI to decide when to call it."""
129	# Implementation
130	return json.dumps(result)
131	```
132
133	#### Resource Generation
134
135	For each resource:
136	- URI template with parameters
137	- Read handler that returns structured content
138	- List handler for collections
139
140	#### Configuration
141
142	Generate `.env.example` with all required environment variables:
143	```env
144	# Required
145	API_KEY=your_api_key_here
146	API_BASE_URL=https://api.example.com
147
148	# Optional
149	LOG_LEVEL=info
150	CACHE_TTL=300
151	```
152
153	Genera

People who liked this also viewed

6 picks

Type