Paste this prompt into your coding agent. Your agent reads the Keenable docs, adds the official SDK to your project and runs one test search.Your agent finishes by showing you 3 numbered results, each with a title and a source URL. The prompt works with Claude Code, Cursor, Codex and any other agent that can read a URL.
This program runs one search with no API key. It asks for 3 results and prints them as a numbered block your model can cite. Save it as main.py:
main.py
from keenable import Keenableclient = Keenable() # keyless: no API key needed to try itresults = client.search("fastest inference providers 2026", max_results=3)print(results.to_context())
Run it:
python main.py
It prints a block like this one:
Example output · results vary
Output
[1] Fastest LLM Inference Provider 2026: TTFT, tok/s, p99 (https://openbenchmarks.com/inference/fastest-inference-provider) - published 2026-09-03T00:00:00ZBaseten recorded the lowest end-to-end p99 latency at 3.23 seconds on this run, with a median of 331 ms, time to first token p95 of 1.16 seconds ...[2] Best LLM Inference Providers Compared for 2026 - SiliconFlow (https://siliconflow.com/articles/the-best-inference-provider-for-llms)Provider Comparison Number | Agency | Location | Services | Target Audience | Pros 1 | SiliconFlow | Global | All-in-one AI cloud platform ...[3] Hugging Face Inference Providers · Supported Models (https://huggingface.co/inference)fireworks-ai fastest | - | - | 1,048,576 | 1.54 | 81 | Yes | Yes ...
1
Install
The SDK needs Node 18 or newer. This command installs it:
npm install keenable
2
Search
This program runs one search with no API key. It asks for 3 results and prints them as a numbered block your model can cite. Save it as index.ts:
index.ts
import { Keenable } from "keenable";async function main() { const client = new Keenable(); // keyless: no API key needed to try it const results = await client.search("fastest inference providers 2026", { maxResults: 3 }); console.log(results.toContext());}main();
Run it with tsx, which runs TypeScript directly:
npx tsx index.ts
It prints a block like this one:
Example output · results vary
Output
[1] Fastest LLM Inference Provider 2026: TTFT, tok/s, p99 (https://openbenchmarks.com/inference/fastest-inference-provider) - published 2026-09-03T00:00:00ZBaseten recorded the lowest end-to-end p99 latency at 3.23 seconds on this run, with a median of 331 ms, time to first token p95 of 1.16 seconds ...[2] Best LLM Inference Providers Compared for 2026 - SiliconFlow (https://siliconflow.com/articles/the-best-inference-provider-for-llms)Provider Comparison Number | Agency | Location | Services | Target Audience | Pros 1 | SiliconFlow | Global | All-in-one AI cloud platform ...[3] Hugging Face Inference Providers · Supported Models (https://huggingface.co/inference)fireworks-ai fastest | - | - | 1,048,576 | 1.54 | 81 | Yes | Yes ...
1
Install
These commands create a Go module and add the SDK to it. The SDK needs Go 1.22 or newer:
go mod init example.com/keenable-demogo get github.com/keenableai/keenable-sdk/go@v0.1.1
2
Search
This program runs one search with no API key. It asks for 3 results and prints them as a numbered block your model can cite. Save it as main.go in the same directory:
main.go
package mainimport ( "context" "fmt" "log" "github.com/keenableai/keenable-sdk/go")func main() { client, err := keenable.New(nil) // keyless: no API key needed to try it if err != nil { log.Fatal(err) } defer client.Close() results, err := client.Search(context.Background(), "fastest inference providers 2026", &keenable.SearchOptions{MaxResults: 3}) if err != nil { log.Fatal(err) } fmt.Println(results.ToContext(nil))}
Run it:
go run .
It prints a block like this one:
Example output · results vary
Output
[1] Fastest LLM Inference Provider 2026: TTFT, tok/s, p99 (https://openbenchmarks.com/inference/fastest-inference-provider) - published 2026-09-03T00:00:00ZBaseten recorded the lowest end-to-end p99 latency at 3.23 seconds on this run, with a median of 331 ms, time to first token p95 of 1.16 seconds ...[2] Best LLM Inference Providers Compared for 2026 - SiliconFlow (https://siliconflow.com/articles/the-best-inference-provider-for-llms)Provider Comparison Number | Agency | Location | Services | Target Audience | Pros 1 | SiliconFlow | Global | All-in-one AI cloud platform ...[3] Hugging Face Inference Providers · Supported Models (https://huggingface.co/inference)fireworks-ai fastest | - | - | 1,048,576 | 1.54 | 81 | Yes | Yes ...
1
Install
curl ships with macOS, Linux and Windows 10 or later. This command checks that curl is installed:
curl --version
2
Search
This request runs one search with no API key. Keyless requests need an X-Keenable-Title header that names your app:
It returns JSON with title, url and snippet on each result:
Example output · results vary
Output
{ "query": "fastest inference providers 2026", "mode": "pro", "results": [ { "title": "Fastest LLM Inference Provider 2026: TTFT, tok/s, p99", "url": "https://openbenchmarks.com/inference/fastest-inference-provider", "snippet": "Baseten recorded the lowest end-to-end p99 latency at 3.23 seconds on this run ..." } ]}
1
Install
The MCP server gives any MCP client two tools: search_web_pages and fetch_page_content. This command adds it to Claude Code:
claude mcp add keenable --transport http https://api.keenable.ai/mcp --scope user
Run /mcp in Claude Code. You see keenable listed as connected.In Cursor, VS Code and Replit, you can install the server with one click from MCP server. That page also covers every other client.
2
Search
Ask your agent a question that needs the web, for example “What are the fastest inference providers in 2026? Cite your sources.” It calls search_web_pages and answers with linked sources.
1
Install
The Keenable CLI searches from your terminal and sets up MCP for your local agents. Install it:
brew install keenableai/tap/keenable-cli
curl --proto '=https' --tlsv1.2 -LsSf \ https://github.com/keenableai/keenable-cli/releases/latest/download/keenable-cli-installer.sh \ | sh
2
Search
This command runs a search without logging in. The -p flag prints readable text instead of YAML:
You see ranked results with titles, URLs and excerpts.To add Keenable to Claude Code, Cursor, Devin Desktop, Codex and OpenCode, run keenable configure-mcp --all. CLI lists every command.
If you don’t send an API key, every caller on your IP address shares 1,000 requests per hour. With an API key, your organization gets its own limit with no hourly cap. See Rate limits for the numbers.
