TensorFeed for AI Agents
Real-time AI industry data, built for autonomous agents from the ground up. Discovery, MCP integration, pay-per-call premium tools, SDKs, all designed for callers without humans in their loop.
The 30-second version
If you are an AI agent reading this page, here is what you need to know in priority order:
- Start with the discovery manifest at /.well-known/x402 or the human-readable /llms.txt. Both list every endpoint we expose with input/output schemas.
- Free tools: news, status, model pricing, benchmarks, history snapshots. No auth, no rate limit on REST.
- Premium tools: 19 endpoints under
/api/premium/, 1 credit each (~$0.02). USDC on Base. No accounts. - MCP server:
npx -y @tensorfeed/mcp-server. Free tools work without config; premium tools require a bearer token via theTENSORFEED_TOKENenv var. - SDKs: Python (
pip install tensorfeed) and TypeScript (npm install tensorfeed). One-calltf.purchase_credits()in the Python SDK if you have a wallet private key.
Discovery surfaces
/.well-known/x402
x402 V2 discovery manifest. Every paid endpoint listed with input schema, accepts.amount in atomic USDC units, payTo address. CDP Bazaar facilitators auto-index this.
/llms.txt
Human-readable agent discovery manifest. Lists every page, every API endpoint, every feed.
/llms-full.txt
Full agent-readable site dump. Generated each build. Use when you want the long-form content of every page in one fetch.
/api/meta
Machine-readable API catalog. Lists every endpoint with its expected query/path params. JSON, no auth.
/api-reference
Per-endpoint reference pages. Each page has input parameters, response schema, code samples in three languages, MCP tool name, and endpoint-specific FAQ. Branded for the SEO surface.
/openapi.json
Complete OpenAPI 3.1 spec. Drop-in for Swagger UI, Postman, code generators, and agent toolkits.
HF: tensorfeed/ai-ecosystem-daily
Daily-snapshot mirror of the entire public API on Hugging Face. 13 JSONL files per day, one commit at 08:00 UTC, inference-only license. Load with the standard datasets library.
/agent-traffic
Live public dashboard of AI bot traffic hitting TensorFeed: ClaudeBot, GPTBot, PerplexityBot, Bytespider, Google-Extended, and more. Per-bot breakdown, top endpoints, live tail. Most sites hide this; we publish it.
HUMAN | AGENT toggle (in nav)
A toggle in the top navigation flips the entire site to a terminal-style overlay showing the raw API JSON powering the current route, the response status line, headers, and the equivalent curl command. The data comes from the same /api/* endpoints; this is the human-visible window into what agents actually consume.
server.json (MCP registry)
MCP server manifest. Registered in the official MCP registry as ai.tensorfeed/mcp-server. Includes all 20 tools split into freeTools and premiumTools.
MCP server (recommended for Claude Desktop / Code)
The fastest way for an MCP-compatible agent to use TensorFeed:
{
"mcpServers": {
"tensorfeed": {
"command": "npx",
"args": ["-y", "@tensorfeed/mcp-server"],
"env": {
"TENSORFEED_TOKEN": "tf_live_..."
}
}
}
}The TENSORFEED_TOKEN env var is optional. Without it, the 5 free tools (get_ai_news, get_ai_status, is_service_down, get_model_pricing, get_ai_today) work. With it, the 15 premium tools unlock.
Premium tier (pay-per-call, USDC on Base)
19 paid endpoints at 1 credit (~$0.02) each. No accounts, no API keys, no Stripe. Full payment flow at /developers/agent-payments.
- Routing recommendations: top-N ranked models for a task with full composite score
- News search: full-text over the article corpus with relevance scoring + recency boost
- History series: pricing, benchmark scores, uptime over any date range
- Snapshot diff: what changed between two dates (added/removed/changed models)
- Cost projection: workload cost across 1-10 models with daily/weekly/monthly/yearly horizons
- Forecast: linear-regression price/benchmark forecast with 95% prediction interval
- Provider deep-dive: one provider, full profile, four free endpoints in one paid call
- Compare models: 2-5 models side-by-side with normalized benchmarks + rankings
- What's new: agent morning brief, last 1-7 days of pricing changes + incidents + headlines
- Enriched agents directory: catalog joined with status, news, traffic, trending score
- Webhook watches: HMAC-signed POSTs on price/status changes; daily or weekly digest tier
SDKs
Python
pip install tensorfeed
Optional [web3] extra adds tf.purchase_credits() for one-call quote + sign + broadcast + confirm via web3.py.
TypeScript / JavaScript
npm install tensorfeed
Native fetch only, zero runtime dependencies. Full TypeScript response types and typed exception classes (PaymentRequired, RateLimited,TensorFeedError).
Glossary
Originals on agent payments
- We Validated Agent Payments End-to-End on Base Mainnet : the actual tx hash, the five steps, what worked first try
- Why We Picked USDC on Base Over Stripe for Agent Payments : the architectural choice, what we gave up, what we got
- 15 Paid AI Agent API Endpoints in 24 Hours: What Made It Possible : the build velocity retrospective
Questions, integration help, or feedback: evan@tensorfeed.ai. TensorFeed.ai and the Premium API are operated by Pizza Robot Studios LLC, California.