llmbenchmark-mcp 1.0.0__tar.gz

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.
@@ -0,0 +1,11 @@
1
+ {
2
+ "mcpServers": {
3
+ "llmbenchmark": {
4
+ "url": "https://api.llmbenchmark.ai/mcp"
5
+ },
6
+ "llmbenchmark-local": {
7
+ "command": "uvx",
8
+ "args": ["llmbenchmark-mcp"]
9
+ }
10
+ }
11
+ }
@@ -0,0 +1,4 @@
1
+ dist/
2
+ __pycache__/
3
+ .venv/
4
+ *.egg-info/
@@ -0,0 +1,6 @@
1
+ FROM python:3.12-slim
2
+ WORKDIR /app
3
+ COPY . /app
4
+ RUN pip install --no-cache-dir .
5
+ ENV PYTHONUNBUFFERED=1
6
+ ENTRYPOINT ["llmbenchmark-mcp"]
@@ -0,0 +1,21 @@
1
+ MIT License
2
+
3
+ Copyright (c) 2026 TAKE GROUP sp. z o.o.
4
+
5
+ Permission is hereby granted, free of charge, to any person obtaining a copy
6
+ of this software and associated documentation files (the "Software"), to deal
7
+ in the Software without restriction, including without limitation the rights
8
+ to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9
+ copies of the Software, and to permit persons to whom the Software is
10
+ furnished to do so, subject to the following conditions:
11
+
12
+ The above copyright notice and this permission notice shall be included in all
13
+ copies or substantial portions of the Software.
14
+
15
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16
+ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17
+ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18
+ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19
+ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21
+ SOFTWARE.
@@ -0,0 +1,139 @@
1
+ Metadata-Version: 2.5
2
+ Name: llmbenchmark-mcp
3
+ Version: 1.0.0
4
+ Summary: Which LLM for which task: benchmark scores and API prices for 400+ models as an MCP server. Free, no API key.
5
+ Project-URL: Homepage, https://llmbenchmark.ai/api?utm_source=pypi&utm_medium=package&utm_campaign=mcp
6
+ Project-URL: Source, https://github.com/Take-Group/llmbenchmark-mcp
7
+ Author-email: "LLMBenchmark.ai" <hello@llmbenchmark.ai>
8
+ License-Expression: MIT
9
+ License-File: LICENSE
10
+ Keywords: chatgpt,claude,cursor,llm api pricing,llm benchmark,llm leaderboard,mcp server,model comparison
11
+ Classifier: Programming Language :: Python :: 3
12
+ Classifier: Topic :: Software Development
13
+ Requires-Python: >=3.10
14
+ Description-Content-Type: text/markdown
15
+
16
+ # LLMBenchmark.ai MCP server
17
+
18
+ Ask "which model should I use for this task?" and get an answer from benchmark scores and API prices of 400+ LLMs, updated daily.
19
+
20
+ - Free. No API key. 500 requests per IP per UTC day.
21
+ - Remote server, Streamable HTTP: `https://api.llmbenchmark.ai/mcp`
22
+ - Deterministic: no LLM call inside, the same task and the same data return the same answer.
23
+ - Every answer carries the data date, the source ("LLMBenchmark.ai") and links to the pages behind the numbers.
24
+
25
+ Docs: https://llmbenchmark.ai/api
26
+
27
+ ## Install
28
+
29
+ **Claude Code**
30
+
31
+ ```bash
32
+ claude mcp add --transport http llmbenchmark https://api.llmbenchmark.ai/mcp
33
+ ```
34
+
35
+ **Cursor** (`.cursor/mcp.json`)
36
+
37
+ ```json
38
+ {
39
+ "mcpServers": {
40
+ "llmbenchmark": { "url": "https://api.llmbenchmark.ai/mcp" }
41
+ }
42
+ }
43
+ ```
44
+
45
+ **Codex** (`~/.codex/config.toml`)
46
+
47
+ ```toml
48
+ [mcp_servers.llmbenchmark]
49
+ url = "https://api.llmbenchmark.ai/mcp"
50
+ ```
51
+
52
+ **Claude Desktop** (`claude_desktop_config.json`, through the `mcp-remote` bridge)
53
+
54
+ ```json
55
+ {
56
+ "mcpServers": {
57
+ "llmbenchmark": {
58
+ "command": "npx",
59
+ "args": ["-y", "mcp-remote", "https://api.llmbenchmark.ai/mcp"]
60
+ }
61
+ }
62
+ }
63
+ ```
64
+
65
+ **Install as a local server (stdio)**
66
+
67
+ The package proxies to the hosted server, so clients that only start local commands can use it too. Needs [uv](https://docs.astral.sh/uv/).
68
+
69
+ ```bash
70
+ uvx llmbenchmark-mcp
71
+ ```
72
+
73
+ Claude Desktop (`claude_desktop_config.json`):
74
+
75
+ ```json
76
+ {
77
+ "mcpServers": {
78
+ "llmbenchmark": {
79
+ "command": "uvx",
80
+ "args": ["llmbenchmark-mcp"]
81
+ }
82
+ }
83
+ }
84
+ ```
85
+
86
+ Docker, built from this repository:
87
+
88
+ ```bash
89
+ docker build -t llmbenchmark-mcp . && docker run -i --rm llmbenchmark-mcp
90
+ ```
91
+
92
+ Set `LLMBENCHMARK_MCP_URL` to point the proxy at another endpoint.
93
+
94
+ ## Tools
95
+
96
+ | Tool | What it does |
97
+ |---|---|
98
+ | `recommend_model` | Task text plus optional limits (`max_output_price_per_m`, `min_context_tokens`, `open_weights_only`, `top_k`) in, ranked models with scores, prices and a one-line reason out. |
99
+ | `get_leaderboard` | Top models on one board: overall, coding, agentic, reasoning, math, knowledge, design. |
100
+ | `compare_models` | Two models side by side on every board where both were measured. |
101
+ | `get_model` | One model: maker, prices, context, open weights, score and rank per board. |
102
+ | `get_prices` | API list prices per million tokens, cheapest first, by model or maker. |
103
+ | `list_categories` | The categories, the board behind each and the task keywords that select it. |
104
+
105
+ `recommend_model` maps the task to a category with a fixed keyword map (printed at https://llmbenchmark.ai/api#how-recommend-works), drops models that break your limits and ranks the rest on that category's board. If nothing fits your limits it says so and returns the nearest options.
106
+
107
+ ## Example prompts
108
+
109
+ - "Which model should I use to refactor a React app under $5 per million output tokens?"
110
+ - "What is the best open-weight model for math with at least a 128k context window?"
111
+ - "Compare Claude Opus and GPT-6 Astra for coding."
112
+ - "Which models are cheapest on output price from DeepSeek?"
113
+
114
+ ## CLI
115
+
116
+ ```bash
117
+ llmbenchmark-mcp recommend "refactor a React app" --max-price 5 --open-weights
118
+ llmbenchmark-mcp tools
119
+ llmbenchmark-mcp config
120
+ ```
121
+
122
+ `recommend` prints the `recommend_model` answer, `tools` lists the tool names, `config` prints the remote JSON config for Claude or Cursor. With no arguments the command runs the stdio server.
123
+
124
+ ## REST
125
+
126
+ The same answers over plain HTTP:
127
+
128
+ ```bash
129
+ curl "https://api.llmbenchmark.ai/v1/recommend?task=refactor+a+large+React+codebase&max_output_price_per_m=5"
130
+ ```
131
+
132
+ ## License
133
+
134
+ - This README and the configuration files in this repository: MIT, see `LICENSE`.
135
+ - The data the server returns: CC BY 4.0. Credit "LLMBenchmark.ai" with a link to https://llmbenchmark.ai. Each score names its source board and the day it was read; the boards keep their own terms.
136
+
137
+ Run by [Take Group](https://takegroup.pl).
138
+
139
+ <!-- mcp-name: ai.llmbenchmark/llmbenchmark -->
@@ -0,0 +1,124 @@
1
+ # LLMBenchmark.ai MCP server
2
+
3
+ Ask "which model should I use for this task?" and get an answer from benchmark scores and API prices of 400+ LLMs, updated daily.
4
+
5
+ - Free. No API key. 500 requests per IP per UTC day.
6
+ - Remote server, Streamable HTTP: `https://api.llmbenchmark.ai/mcp`
7
+ - Deterministic: no LLM call inside, the same task and the same data return the same answer.
8
+ - Every answer carries the data date, the source ("LLMBenchmark.ai") and links to the pages behind the numbers.
9
+
10
+ Docs: https://llmbenchmark.ai/api
11
+
12
+ ## Install
13
+
14
+ **Claude Code**
15
+
16
+ ```bash
17
+ claude mcp add --transport http llmbenchmark https://api.llmbenchmark.ai/mcp
18
+ ```
19
+
20
+ **Cursor** (`.cursor/mcp.json`)
21
+
22
+ ```json
23
+ {
24
+ "mcpServers": {
25
+ "llmbenchmark": { "url": "https://api.llmbenchmark.ai/mcp" }
26
+ }
27
+ }
28
+ ```
29
+
30
+ **Codex** (`~/.codex/config.toml`)
31
+
32
+ ```toml
33
+ [mcp_servers.llmbenchmark]
34
+ url = "https://api.llmbenchmark.ai/mcp"
35
+ ```
36
+
37
+ **Claude Desktop** (`claude_desktop_config.json`, through the `mcp-remote` bridge)
38
+
39
+ ```json
40
+ {
41
+ "mcpServers": {
42
+ "llmbenchmark": {
43
+ "command": "npx",
44
+ "args": ["-y", "mcp-remote", "https://api.llmbenchmark.ai/mcp"]
45
+ }
46
+ }
47
+ }
48
+ ```
49
+
50
+ **Install as a local server (stdio)**
51
+
52
+ The package proxies to the hosted server, so clients that only start local commands can use it too. Needs [uv](https://docs.astral.sh/uv/).
53
+
54
+ ```bash
55
+ uvx llmbenchmark-mcp
56
+ ```
57
+
58
+ Claude Desktop (`claude_desktop_config.json`):
59
+
60
+ ```json
61
+ {
62
+ "mcpServers": {
63
+ "llmbenchmark": {
64
+ "command": "uvx",
65
+ "args": ["llmbenchmark-mcp"]
66
+ }
67
+ }
68
+ }
69
+ ```
70
+
71
+ Docker, built from this repository:
72
+
73
+ ```bash
74
+ docker build -t llmbenchmark-mcp . && docker run -i --rm llmbenchmark-mcp
75
+ ```
76
+
77
+ Set `LLMBENCHMARK_MCP_URL` to point the proxy at another endpoint.
78
+
79
+ ## Tools
80
+
81
+ | Tool | What it does |
82
+ |---|---|
83
+ | `recommend_model` | Task text plus optional limits (`max_output_price_per_m`, `min_context_tokens`, `open_weights_only`, `top_k`) in, ranked models with scores, prices and a one-line reason out. |
84
+ | `get_leaderboard` | Top models on one board: overall, coding, agentic, reasoning, math, knowledge, design. |
85
+ | `compare_models` | Two models side by side on every board where both were measured. |
86
+ | `get_model` | One model: maker, prices, context, open weights, score and rank per board. |
87
+ | `get_prices` | API list prices per million tokens, cheapest first, by model or maker. |
88
+ | `list_categories` | The categories, the board behind each and the task keywords that select it. |
89
+
90
+ `recommend_model` maps the task to a category with a fixed keyword map (printed at https://llmbenchmark.ai/api#how-recommend-works), drops models that break your limits and ranks the rest on that category's board. If nothing fits your limits it says so and returns the nearest options.
91
+
92
+ ## Example prompts
93
+
94
+ - "Which model should I use to refactor a React app under $5 per million output tokens?"
95
+ - "What is the best open-weight model for math with at least a 128k context window?"
96
+ - "Compare Claude Opus and GPT-6 Astra for coding."
97
+ - "Which models are cheapest on output price from DeepSeek?"
98
+
99
+ ## CLI
100
+
101
+ ```bash
102
+ llmbenchmark-mcp recommend "refactor a React app" --max-price 5 --open-weights
103
+ llmbenchmark-mcp tools
104
+ llmbenchmark-mcp config
105
+ ```
106
+
107
+ `recommend` prints the `recommend_model` answer, `tools` lists the tool names, `config` prints the remote JSON config for Claude or Cursor. With no arguments the command runs the stdio server.
108
+
109
+ ## REST
110
+
111
+ The same answers over plain HTTP:
112
+
113
+ ```bash
114
+ curl "https://api.llmbenchmark.ai/v1/recommend?task=refactor+a+large+React+codebase&max_output_price_per_m=5"
115
+ ```
116
+
117
+ ## License
118
+
119
+ - This README and the configuration files in this repository: MIT, see `LICENSE`.
120
+ - The data the server returns: CC BY 4.0. Credit "LLMBenchmark.ai" with a link to https://llmbenchmark.ai. Each score names its source board and the day it was read; the boards keep their own terms.
121
+
122
+ Run by [Take Group](https://takegroup.pl).
123
+
124
+ <!-- mcp-name: ai.llmbenchmark/llmbenchmark -->
@@ -0,0 +1,12 @@
1
+ # Example prompts
2
+
3
+ Ask these in any client with the LLMBenchmark.ai server connected.
4
+
5
+ - "Which model should I use to refactor a React app under $5 per million output tokens?"
6
+ - "What is the best open-weight model for math with at least a 128k context window?"
7
+ - "Compare Claude Opus and GPT-6 Astra for coding."
8
+ - "Which models are cheapest on output price from DeepSeek?"
9
+ - "Which model is best for agentic tool use with a 200k context window?"
10
+ - "Show the top 10 models on the reasoning leaderboard."
11
+ - "What does Gemini cost per million input and output tokens compared with Claude Sonnet?"
12
+ - "Which categories can you rank models on, and which task keywords pick each one?"
@@ -0,0 +1 @@
1
+ {"$schema": "https://glama.ai/mcp/schemas/server.json", "maintainers": ["slugc"]}
@@ -0,0 +1,185 @@
1
+ #!/usr/bin/env python3
2
+ """LLMBenchmark.ai MCP server: stdio proxy and small CLI over the hosted server (stdlib only)."""
3
+ import argparse
4
+ import json
5
+ import os
6
+ import sys
7
+ import urllib.error
8
+ import urllib.request
9
+
10
+ VERSION = "1.0.0"
11
+ URL = os.environ.get("LLMBENCHMARK_MCP_URL") or "https://api.llmbenchmark.ai/mcp"
12
+ PROTOCOLS = ("2025-06-18", "2025-03-26", "2024-11-05")
13
+ USER_AGENT = "llmbenchmark-mcp/" + VERSION
14
+
15
+ CONFIG = {"mcpServers": {"llmbenchmark": {"url": "https://api.llmbenchmark.ai/mcp"}}}
16
+
17
+
18
+ class Client:
19
+ """Talks to the hosted Streamable HTTP server; handles JSON and SSE replies and the session id."""
20
+
21
+ def __init__(self):
22
+ self.session = None
23
+ self.next_id = 0
24
+ self.started = False
25
+ self.tools = None
26
+
27
+ def _post(self, payload):
28
+ headers = {
29
+ "Accept": "application/json, text/event-stream",
30
+ "Content-Type": "application/json",
31
+ "User-Agent": USER_AGENT,
32
+ }
33
+ if self.session:
34
+ headers["Mcp-Session-Id"] = self.session
35
+ req = urllib.request.Request(URL, json.dumps(payload).encode(), headers, method="POST")
36
+ try:
37
+ with urllib.request.urlopen(req, timeout=60) as resp:
38
+ self.session = resp.headers.get("Mcp-Session-Id") or self.session
39
+ return resp.read().decode(), resp.headers.get("Content-Type", "")
40
+ except urllib.error.HTTPError as e:
41
+ raise RuntimeError("HTTP %s: %s" % (e.code, e.read().decode(errors="replace")[:300]))
42
+ except urllib.error.URLError as e:
43
+ raise RuntimeError("Cannot reach %s: %s" % (URL, e.reason))
44
+
45
+ @staticmethod
46
+ def _parse(body, ctype, rid):
47
+ if "text/event-stream" in ctype:
48
+ for line in body.splitlines():
49
+ if line.startswith("data:"):
50
+ try:
51
+ msg = json.loads(line[5:].strip())
52
+ except ValueError:
53
+ continue
54
+ if isinstance(msg, dict) and msg.get("id") == rid:
55
+ return msg
56
+ raise RuntimeError("No JSON-RPC answer in stream")
57
+ return json.loads(body)
58
+
59
+ def rpc(self, method, params=None):
60
+ self.next_id += 1
61
+ rid = self.next_id
62
+ body, ctype = self._post({"jsonrpc": "2.0", "id": rid, "method": method, "params": params or {}})
63
+ msg = self._parse(body, ctype, rid)
64
+ if "error" in msg:
65
+ raise RuntimeError(msg["error"].get("message", str(msg["error"])))
66
+ return msg.get("result", {})
67
+
68
+ def start(self):
69
+ if self.started:
70
+ return
71
+ self.rpc("initialize", {
72
+ "protocolVersion": PROTOCOLS[0],
73
+ "capabilities": {},
74
+ "clientInfo": {"name": "llmbenchmark-mcp", "version": VERSION},
75
+ })
76
+ self._post({"jsonrpc": "2.0", "method": "notifications/initialized"})
77
+ self.started = True
78
+
79
+ def list_tools(self):
80
+ if self.tools is None:
81
+ self.start()
82
+ self.tools = self.rpc("tools/list").get("tools", [])
83
+ return self.tools
84
+
85
+ def call(self, name, args):
86
+ self.start()
87
+ return self.rpc("tools/call", {"name": name, "arguments": args or {}})
88
+
89
+
90
+ def result_text(res):
91
+ return "\n".join(c.get("text", "") for c in res.get("content", []) if c.get("type") == "text")
92
+
93
+
94
+ class Server:
95
+ def __init__(self):
96
+ self.client = Client()
97
+
98
+ def handle(self, msg):
99
+ method, rid, params = msg.get("method"), msg.get("id"), msg.get("params") or {}
100
+ if rid is None:
101
+ return None # notification, no reply
102
+ if method == "initialize":
103
+ want = params.get("protocolVersion")
104
+ result = {
105
+ "protocolVersion": want if want in PROTOCOLS else PROTOCOLS[0],
106
+ "capabilities": {"tools": {"listChanged": False}},
107
+ "serverInfo": {"name": "llmbenchmark", "title": "LLMBenchmark.ai", "version": VERSION},
108
+ }
109
+ elif method == "ping":
110
+ result = {}
111
+ elif method == "tools/list":
112
+ try:
113
+ result = {"tools": self.client.list_tools()}
114
+ except Exception as e:
115
+ return {"jsonrpc": "2.0", "id": rid, "error": {"code": -32603, "message": str(e)}}
116
+ elif method == "tools/call":
117
+ try:
118
+ result = self.client.call(params.get("name", ""), params.get("arguments"))
119
+ except Exception as e:
120
+ self.client = Client()
121
+ result = {"content": [{"type": "text", "text": "LLMBenchmark.ai is unreachable: %s" % e}], "isError": True}
122
+ else:
123
+ return {"jsonrpc": "2.0", "id": rid, "error": {"code": -32601, "message": "Method not found: %s" % method}}
124
+ return {"jsonrpc": "2.0", "id": rid, "result": result}
125
+
126
+
127
+ def serve():
128
+ """MCP stdio transport: one JSON-RPC message per line on stdin/stdout, logs on stderr."""
129
+ server = Server()
130
+ for line in sys.stdin:
131
+ line = line.strip()
132
+ if not line:
133
+ continue
134
+ msg = None
135
+ try:
136
+ msg = json.loads(line)
137
+ reply = server.handle(msg) if isinstance(msg, dict) else None
138
+ except ValueError:
139
+ reply = {"jsonrpc": "2.0", "id": None, "error": {"code": -32700, "message": "Parse error"}}
140
+ except Exception as e:
141
+ print("Error: %s" % e, file=sys.stderr)
142
+ reply = {"jsonrpc": "2.0", "id": msg.get("id") if isinstance(msg, dict) else None,
143
+ "error": {"code": -32603, "message": str(e)}}
144
+ if reply is not None:
145
+ sys.stdout.write(json.dumps(reply) + "\n")
146
+ sys.stdout.flush()
147
+
148
+
149
+ def main():
150
+ p = argparse.ArgumentParser(prog="llmbenchmark-mcp", description="LLMBenchmark.ai MCP server and CLI. No args: stdio server.")
151
+ p.add_argument("--version", action="version", version="llmbenchmark-mcp " + VERSION)
152
+ sub = p.add_subparsers(dest="cmd")
153
+ r = sub.add_parser("recommend", help="print the recommend_model answer for a task")
154
+ r.add_argument("task")
155
+ r.add_argument("--max-price", type=float, help="max output price per million tokens")
156
+ r.add_argument("--open-weights", action="store_true", help="open-weight models only")
157
+ sub.add_parser("tools", help="print the tool names")
158
+ sub.add_parser("config", help="print the remote JSON config for Claude/Cursor")
159
+ args = p.parse_args()
160
+ if args.cmd is None:
161
+ return serve()
162
+ if args.cmd == "config":
163
+ print(json.dumps(CONFIG, indent=2))
164
+ return
165
+ try:
166
+ c = Client()
167
+ if args.cmd == "tools":
168
+ print("\n".join(t["name"] for t in c.list_tools()))
169
+ else:
170
+ a = {"task": args.task}
171
+ if args.max_price is not None:
172
+ a["max_output_price_per_m"] = args.max_price
173
+ if args.open_weights:
174
+ a["open_weights_only"] = True
175
+ res = c.call("recommend_model", a)
176
+ print(result_text(res))
177
+ if res.get("isError"):
178
+ sys.exit(1)
179
+ except Exception as e:
180
+ print("Error: %s" % e, file=sys.stderr)
181
+ sys.exit(1)
182
+
183
+
184
+ if __name__ == "__main__":
185
+ main()
@@ -0,0 +1,11 @@
1
+ {
2
+ "mcpServers": {
3
+ "llmbenchmark": {
4
+ "url": "https://api.llmbenchmark.ai/mcp"
5
+ },
6
+ "llmbenchmark-local": {
7
+ "command": "uvx",
8
+ "args": ["llmbenchmark-mcp"]
9
+ }
10
+ }
11
+ }
@@ -0,0 +1,29 @@
1
+ [project]
2
+ name = "llmbenchmark-mcp"
3
+ version = "1.0.0"
4
+ description = "Which LLM for which task: benchmark scores and API prices for 400+ models as an MCP server. Free, no API key."
5
+ readme = "README.md"
6
+ license = "MIT"
7
+ license-files = ["LICENSE"]
8
+ requires-python = ">=3.10"
9
+ authors = [{ name = "LLMBenchmark.ai", email = "hello@llmbenchmark.ai" }]
10
+ keywords = ["llm benchmark", "llm leaderboard", "llm api pricing", "mcp server", "claude", "chatgpt", "cursor", "model comparison"]
11
+ classifiers = ["Programming Language :: Python :: 3", "Topic :: Software Development"]
12
+ dependencies = []
13
+
14
+ [project.urls]
15
+ Homepage = "https://llmbenchmark.ai/api?utm_source=pypi&utm_medium=package&utm_campaign=mcp"
16
+ Source = "https://github.com/Take-Group/llmbenchmark-mcp"
17
+
18
+ [project.scripts]
19
+ llmbenchmark-mcp = "llmbenchmark_mcp:main"
20
+
21
+ [build-system]
22
+ requires = ["hatchling>=1.27"]
23
+ build-backend = "hatchling.build"
24
+
25
+ [tool.hatch.build.targets.wheel.force-include]
26
+ "llmbenchmark_mcp.py" = "llmbenchmark_mcp.py"
27
+
28
+ [tool.hatch.build.targets.sdist]
29
+ include = ["llmbenchmark_mcp.py", "README.md", "LICENSE", "mcp.json", "Dockerfile", "glama.json", "server.json", "examples", ".cursor"]
@@ -0,0 +1,27 @@
1
+ {
2
+ "$schema": "https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json",
3
+ "name": "ai.llmbenchmark/llmbenchmark",
4
+ "title": "LLMBenchmark.ai",
5
+ "description": "Which LLM for which task: benchmark scores and API prices for 400+ models. Free, no key.",
6
+ "version": "1.0.0",
7
+ "websiteUrl": "https://llmbenchmark.ai/api",
8
+ "repository": {
9
+ "url": "https://github.com/Take-Group/llmbenchmark-mcp",
10
+ "source": "github"
11
+ },
12
+ "remotes": [
13
+ {
14
+ "type": "streamable-http",
15
+ "url": "https://api.llmbenchmark.ai/mcp"
16
+ }
17
+ ],
18
+ "icons": [
19
+ {
20
+ "src": "https://llmbenchmark.ai/apple-touch-icon.png",
21
+ "mimeType": "image/png",
22
+ "sizes": [
23
+ "180x180"
24
+ ]
25
+ }
26
+ ]
27
+ }