LIVE
ANTHROPICOpus 4.7 benchmarks published2m ago
CLAUDEOK142ms
OPUS 4.7$15 / $75per Mtok
CHATGPTOK89ms
HACKERNEWSWhy has not AI improved design quality the way it improved dev speed?14m ago
MMLU-PROleader Opus 4.788.4
GEMINIDEGRADED312ms
MISTRALMistral Medium 3 released6m ago
GPT-4o$5 / $15per Mtok
ARXIVCompositional reasoning in LRMs22m ago
BEDROCKOK178ms
GEMINI 2.5$3.50 / $10.50per Mtok
THE VERGEFrontier Model Forum expansion announced38m ago
SWE-BENCHleader Claude Opus 4.772.1%
MISTRALOK104ms
ANTHROPICOpus 4.7 benchmarks published2m ago
CLAUDEOK142ms
OPUS 4.7$15 / $75per Mtok
CHATGPTOK89ms
HACKERNEWSWhy has not AI improved design quality the way it improved dev speed?14m ago
MMLU-PROleader Opus 4.788.4
GEMINIDEGRADED312ms
MISTRALMistral Medium 3 released6m ago
GPT-4o$5 / $15per Mtok
ARXIVCompositional reasoning in LRMs22m ago
BEDROCKOK178ms
GEMINI 2.5$3.50 / $10.50per Mtok
THE VERGEFrontier Model Forum expansion announced38m ago
SWE-BENCHleader Claude Opus 4.772.1%
MISTRALOK104ms
All use cases

Agent payments integration

Make your AI agent pay TensorFeed (or your own pay-per-call API) in USDC on Base mainnet. End-to-end SDK paths, MCP integration, and the validate-and-charge contract for sister-site Workers. No accounts, no API keys, no Stripe.

Path 1: One-call purchase via the Python SDK

The shortest path for an agent that already has a private key for a Base wallet:

# pip install 'tensorfeed[web3]'
from tensorfeed import TensorFeed

tf = TensorFeed()
result = tf.purchase_credits(
    amount_usd=1.0,
    private_key="0x...",   # never hardcode; read from env or secret manager
)
# Token is auto-stored on tf. Premium calls just work.

rec = tf.routing(task="code", top_n=3)
print(tf.balance())  # ~49 credits remaining

Under the hood this calls /api/payment/buy-credits, signs and broadcasts the USDC transfer via web3.py, then calls /api/payment/confirm. End-to-end takes 5-10 seconds the first time; subsequent premium calls are 50ms.

Path 2: Manual flow (any language)

If you do not want a web3 dependency in your agent, do the three steps yourself:

# 1. Quote
curl -X POST https://tensorfeed.ai/api/payment/buy-credits \
  -H "Content-Type: application/json" \
  -d '{"amount_usd": 1.0}'
# -> {"wallet": "0x549...", "memo": "tf-...", "credits": 50, ...}

# 2. Send USDC on Base to the wallet using your wallet of choice
#    (Rabby, Coinbase Wallet, MetaMask on Base, AgentKit, etc.)

# 3. Confirm
curl -X POST https://tensorfeed.ai/api/payment/confirm \
  -H "Content-Type: application/json" \
  -d '{"tx_hash": "0x...", "nonce": "tf-..."}'
# -> {"token": "tf_live_...", "balance": 50}

# 4. Use
curl https://tensorfeed.ai/api/premium/routing?task=code \
  -H "Authorization: Bearer tf_live_..."

Path 3: x402 fallback (no pre-flight)

For one-shot discovery calls where you do not want to pre-commit to credits, hit any premium endpoint without auth. The 402 response includes payment instructions; retry with X-Payment-Tx: 0x... after sending the USDC. Latency is 3-5 seconds (on-chain confirmation) instead of 50ms, so this is best for discovery, not repeat use.

# Discover the endpoint
curl https://tensorfeed.ai/api/premium/routing
# -> 402 Payment Required + payment instructions

# Send USDC on Base to the wallet from the response, then:
curl https://tensorfeed.ai/api/premium/routing?task=code \
  -H "X-Payment-Tx: 0x..."
# -> data + bearer token in X-Payment-Token header

Path 4: MCP integration (Claude Desktop / Code)

If your agent is an MCP-aware client, drop the TensorFeed MCP server into config. The token lives in an env var; the server reads it on each tool call:

{
  "mcpServers": {
    "tensorfeed": {
      "command": "npx",
      "args": ["-y", "@tensorfeed/mcp-server"],
      "env": { "TENSORFEED_TOKEN": "tf_live_..." }
    }
  }
}

22 tools become available to the model: 5 free (news, status, pricing, etc.) and 17 premium (routing, news search, history series, cost projection, forecasting, provider deep-dive, model comparison, webhook watches, etc.). The model decides which tools to call; you pay 1 credit per premium tool invocation.

Cross-site bundle

TensorFeed exposes /api/internal/validate-and-charge so sister-site Workers (TerminalFeed.io today, future Pizza Robot Studios products tomorrow) can validate the same bearer token and debit credits over HTTP. Server-to- server only, gated by an X-Internal-Auth shared secret, never exposed to agents directly. The result: agents buy credits once, spend them on either site. The architecture is in the build retrospective.

Wallet attestation

Before any agent sends real USDC, it should cross-check the wallet across the four published locations. We publish the address at:

If any of the four disagree, do not send. The address is currently0x549c82e6bfc54bdae9a2073744cbc2af5d1fc6d1on Base mainnet.

If you are building your own agent-payable API

The full architectural argument is in the originals: why USDC on Base over Stripe (link), how validation works end-to-end (link), and a 50-line MCP server reference (link). The TensorFeed source is on GitHub if you want to crib the payment middleware directly.

Recommended endpoints (in priority order)