Live X research, document synthesis, and embeddings. Backed by a private GB10 GPU. Call from curl, any agent framework, or Claude's MCP.
One API call, no card required. Your key arrives in the response.
POST to /v1/run with your tool, params, and API key. Returns a job ID immediately.
GET /v1/jobs/{id} until status is done. Typical doc_synthesis: 30-90s.
# 1. Signup curl -X POST https://api.kyletusing.com/v1/signup \ -H "Content-Type: application/json" \ -d '{"email":"[email protected]"}' # → {"api_key":"ak_..."} # 2. Run a job curl -X POST https://api.kyletusing.com/v1/run \ -H "X-API-Key: ak_YOUR_KEY" \ -H "Content-Type: application/json" \ -d '{ "tool": "doc_synthesis", "params": { "text": "Paste your document here...", "instruction": "Summarize the key points" } }' # → {"job_id":"job_..."} # 3. Poll curl https://api.kyletusing.com/v1/jobs/job_ID \ -H "X-API-Key: ak_YOUR_KEY" # → {"status":"done","result":{"output":"..."}}
Sign up, get a free $10 trial key, then paste any text and ask a question. Runs on private GB10.
Add the stdio MCP server to your Claude config and every tool becomes available inside Claude Desktop or any MCP-compatible client.
Clone the repo or copy mcp_server.py alongside your API key.
Point to the server script with your API key as an env var.
"Research the latest vLLM benchmarks using x_research" , Claude calls the tool, you pay $3.
"mcpServers": { "agent-tools": { "command": "python3", "args": ["/path/to/mcp_server.py"], "env": { "AGENT_TOOLS_API_KEY": "ak_YOUR_KEY", "AGENT_TOOLS_BASE": "https://api.kyletusing.com" } } }
x_research : live X search + synthesis x_research_deep : 30-source deep sweep doc_synthesis : summarize / extract / Q&A embed_index : RAG embeddings (per 1k chunks)