@index9/mcp 1.0.20 → 1.0.22

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (2) hide show
  1. package/README.md +30 -80
  2. package/package.json +5 -6
package/README.md CHANGED
@@ -3,21 +3,13 @@
3
3
  [![npm version](https://badge.fury.io/js/@index9%2Fmcp.svg)](https://badge.fury.io/js/@index9%2Fmcp)
4
4
  [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
5
5
 
6
- [Install](#quick-start) • [Issues](https://github.com/index9-org/mcp/issues)
6
+ MCP server that gives AI assistants up-to-date model data — replacing outdated training knowledge.
7
7
 
8
- MCP server for searching and testing AI models. Data from [OpenRouter](https://openrouter.ai).
8
+ Pricing, context limits, capabilities, and test metrics for 300+ models from [OpenRouter](https://openrouter.ai).
9
9
 
10
- ## Quick Start
10
+ ## Installation
11
11
 
12
- **Zero Config:** Search and Lookup features work immediately after installation. No API keys required.
13
-
14
- ### Cursor
15
-
16
- Open **Cursor Settings** → **MCP** → **Add new global MCP server**
17
-
18
- [![Install MCP Server](https://cursor.com/deeplink/mcp-install-dark.svg)](https://cursor.com/en-US/install-mcp?name=index9&config=eyJjb21tYW5kIjoibnB4IC15IEBpbmRleDkvbWNwIn0%3D)
19
-
20
- Or add manually:
12
+ Add to your MCP client configuration:
21
13
 
22
14
  ```json
23
15
  {
@@ -30,46 +22,26 @@ Or add manually:
30
22
  }
31
23
  ```
32
24
 
33
- ### VS Code
25
+ **Where to add this:**
34
26
 
35
- Add to your **MCP Servers** settings:
27
+ | Client | Location |
28
+ | -------------- | ----------------------------------------------------------------- |
29
+ | Cursor | Settings → MCP → Add new global MCP server |
30
+ | VS Code | Settings → MCP Servers (use `"type": "stdio"`) |
31
+ | Claude Desktop | `~/Library/Application Support/Claude/claude_desktop_config.json` |
32
+ | Claude Code | `claude mcp add index9 -- npx -y @index9/mcp` |
36
33
 
37
- ```json
38
- {
39
- "mcp": {
40
- "servers": {
41
- "index9": {
42
- "type": "stdio",
43
- "command": "npx",
44
- "args": ["-y", "@index9/mcp"]
45
- }
46
- }
47
- }
48
- }
49
- ```
34
+ See [index9.dev](https://index9.dev/#installation) for other supported MCP clients.
50
35
 
51
- ### Claude Desktop
36
+ | Tool | Description | API Key |
37
+ | ------------- | ----------------------------------------------------------------------------------------- | ------- |
38
+ | `find_models` | Search models by natural language or filters (price, context, capabilities). | No |
39
+ | `get_model` | Return full metadata, including pricing, context window, output limits, and capabilities. | No |
40
+ | `test_model` | Send live requests via OpenRouter to compare outputs, latency, and estimated cost. | Yes |
52
41
 
53
- Edit `~/Library/Application Support/Claude/claude_desktop_config.json`:
42
+ ## API Key (Optional)
54
43
 
55
- ```json
56
- {
57
- "mcpServers": {
58
- "index9": {
59
- "command": "npx",
60
- "args": ["-y", "@index9/mcp"]
61
- }
62
- }
63
- }
64
- ```
65
-
66
- See [index9.dev](https://index9.dev/#installation) for Windsurf, Cline, and other clients.
67
-
68
- ## Configuration (Optional)
69
-
70
- The **find_models** and **get_model** tools are free and require no configuration.
71
-
72
- To use the **test_model** tool (for running live API calls), you must provide an OpenRouter API key.
44
+ The `test_model` tool requires an [OpenRouter API key](https://openrouter.ai/keys). Add it to your config:
73
45
 
74
46
  ```json
75
47
  {
@@ -85,46 +57,24 @@ To use the **test_model** tool (for running live API calls), you must provide an
85
57
  }
86
58
  ```
87
59
 
88
- _Your key is never stored. It is only used to make ephemeral requests to OpenRouter for testing purposes._
60
+ Usage is billed to your OpenRouter account. The key is only used for live test requests and is not stored.
89
61
 
90
- ### CLI Tools (Claude Code, etc.)
62
+ ## Auto-Invoke Rule (Optional)
91
63
 
92
- For CLI-based MCP clients, set the environment variable before running:
64
+ Ensures current model data is used instead of outdated training data when model information is relevant.
93
65
 
94
- ```bash
95
- export OPENROUTER_API_KEY="sk-..."
66
+ ```text
67
+ Treat all training-time knowledge about AI models as potentially outdated.
68
+ Whenever a task involves model selection, comparison, pricing, limits, or availability, consult index9 and base decisions on its results.
96
69
  ```
97
70
 
98
- Or add it to your shell profile (`~/.zshrc`, `~/.bashrc`) for persistence.
99
-
100
- ## Tools
71
+ Add to: Cursor Rules, `.windsurfrules`, `CLAUDE.md`, etc.
101
72
 
102
- | Tool | Description | Config Required |
103
- | ------------- | ------------------------------------------------------------------------------------------------------------------------------------ | -------------------- |
104
- | `find_models` | Search and filter 300+ AI models. Use natural language ("fast coding model") or strict filters for price, context, and capabilities. | None |
105
- | `get_model` | Get complete technical specifications. Returns pricing, context windows, max output tokens, and capabilities for any model. | None |
106
- | `test_model` | Run live performance tests. Execute real API calls to multiple models simultaneously to compare latency, token usage, and costs. | `OPENROUTER_API_KEY` |
107
-
108
- ## Usage Examples
109
-
110
- Ask your AI assistant natural questions to find models:
111
-
112
- - "Find a cheap vision model with at least 128k context"
113
- - "What are the best models for coding under $1 per million tokens?"
114
- - "Compare the specs of gpt-4o and claude-3-5-sonnet"
115
- - "Test the latency of haiku vs gemini-flash with a simple 'hello world' prompt"
116
-
117
- ## Tip: Add an Auto-Invoke Rule
118
-
119
- For best results, add a rule so your AI assistant automatically uses index9 when answering model questions:
120
-
121
- ```text
122
- When choosing AI models or comparing pricing/capabilities, use index9 MCP tools
123
- to get current data instead of relying on training knowledge.
124
- ```
73
+ ## Links
125
74
 
126
- Add this to your client's rule system: Cursor Settings → Rules, `.windsurfrules`, `CLAUDE.md`, etc.
75
+ - [Website](https://index9.dev)
76
+ - [Issues](https://github.com/index9-org/mcp/issues)
127
77
 
128
78
  ## License
129
79
 
130
- MIT © [Index9](https://index9.dev)
80
+ MIT
package/package.json CHANGED
@@ -1,8 +1,8 @@
1
1
  {
2
2
  "name": "@index9/mcp",
3
3
  "mcpName": "io.github.index9-org/mcp",
4
- "version": "1.0.20",
5
- "description": "Live Model Registry and performance testing for your AI assistant. Sourced from OpenRouter.",
4
+ "version": "1.0.22",
5
+ "description": "Up-to-date model data for AI assistants. Data from OpenRouter.",
6
6
  "keywords": [
7
7
  "mcp",
8
8
  "model-context-protocol",
@@ -12,7 +12,6 @@
12
12
  "pricing",
13
13
  "capabilities",
14
14
  "context-window",
15
- "real-time",
16
15
  "api",
17
16
  "sdk",
18
17
  "openrouter"
@@ -37,13 +36,13 @@
37
36
  "index9": "./dist/index.js"
38
37
  },
39
38
  "dependencies": {
40
- "@modelcontextprotocol/sdk": "^1.24.3",
39
+ "@modelcontextprotocol/sdk": "^1.25.1",
41
40
  "axios": "^1.13.2",
42
41
  "pino": "^10.1.0",
43
- "zod": "^4.1.13"
42
+ "zod": "^4.2.1"
44
43
  },
45
44
  "devDependencies": {
46
- "@types/node": "^25.0.1",
45
+ "@types/node": "^25.0.3",
47
46
  "pino-pretty": "^13.1.3",
48
47
  "tsx": "^4.21.0",
49
48
  "typescript": "^5.9.3"