AI inference bridge built exclusively for AI agents — no API keys, pay with Solana USDC via x402.
⚠️ Billing note (reasoning models):
max_tokens >= 200 on models that reason by default (thinking is disabled on: glm-4.7-flash, deepseek-flash, deepseek-v4-pro) — a smaller limit can return an empty answer. We retry once automatically with a bigger budget; if the answer is still empty you get an error and a full refund (§2.4).First call — free, copy this one: the leading free model answers with no wallet and no payment. Keep max_tokens >= 200 (a smaller limit can be eaten by reasoning and return an empty answer).
curl https://bridgenode.cc/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{"model":"gpt-oss-20b","messages":[{"role":"user","content":"hello"}],"max_tokens":200}'
Then, for a PAID model, three steps — no API key, no registration. Any x402-capable client or the official SDKs handle the handshake automatically.
POST https://bridgenode.cc/v1/chat/completions — response is 402 Payment Required with a PAYMENT-REQUIRED header (price, payTo, memo, blockhash).TransferChecked + Memo transaction with your wallet. Fee payer is sponsored — the agent only needs USDC.PAYMENT-SIGNATURE header — response is 200 with a PAYMENT-RESPONSE settlement receipt.curl https://bridgenode.cc/v1/chat/completions -H "Content-Type: application/json" -d '{"model":"deepseek-flash","messages":[{"role":"user","content":"hello"}],"max_tokens":200}'
# → 402 Payment Required + PAYMENT-REQUIRED header
# sign the partial transaction, then retry with PAYMENT-SIGNATURE header
# → 200 OK + PAYMENT-RESPONSE header
Live prices: /v1/models. Full flow and SDK examples: /llms-full.txt, /skill.md.
Add BridgeNode to any MCP client — one line (stdio wrapper, @bridgenode/mcp):
claude mcp add bridgenode -s user -- npx -y @bridgenode/mcp@latest
Or use the remote server directly (streamable-http) in claude_desktop_config.json / .cursor/mcp.json / .windsurf/mcp.json:
{
"mcpServers": {
"bridgenode": {
"type": "http",
"url": "https://bridgenode.cc/mcp"
}
}
}
Tools: chat_completions (paid, x402 — accepts OpenAI-style tools for agent tool calling), list_models and get_price_estimate (free).
POST https://bridgenode.cc/v1/chat/completions — chat (x402 paid)GET https://bridgenode.cc/v1/models — models + prices (public)https://bridgenode.cc/mcp — MCP server (x402 paid tools)No payment, no API key, no registration — just send a request with the model id (first request → 200 directly).
gpt-oss-20b, gpt-oss-120b, glm-4.7-flash, glm-4.5-flash, glm-4.6v-flashFree model notes (read before choosing one):
glm-4.7-flash — ⚠️ temporarily unreliable: z.ai free model: slower than the Groq free models — a reply can take up to a minute, and the provider is sometimes overloaded. If it returns an error (rate limit / temporarily overloaded), retry once or switch to gpt-oss-20b, the most reliable free model.glm-4.5-flash — ⚠️ temporarily unreliable: z.ai free model: slower than the Groq free models — a reply can take up to a minute, and the provider is sometimes overloaded. If it returns an error (rate limit / temporarily overloaded), retry once or switch to gpt-oss-20b, the most reliable free model.glm-4.6v-flash: z.ai free model: slower than the Groq free models — a reply can take up to a minute, and the provider is sometimes overloaded. If it returns an error (rate limit / temporarily overloaded), retry once or switch to gpt-oss-20b, the most reliable free model.gpt-oss-20b 🆓 — $0.00/1M in, $0.00/1M outgpt-oss-120b 🆓 — $0.00/1M in, $0.00/1M outglm-4.7-flash 🆓 — $0.00/1M in, $0.00/1M outglm-4.5-flash 🆓 — $0.00/1M in, $0.00/1M outglm-4.6v-flash 🆓 — $0.00/1M in, $0.00/1M outdeepseek-flash — $0.18/1M in, $0.70/1M outglm-4.7-flashx — $0.08/1M in, $0.47/1M outglm-5.2 — $1.64/1M in, $5.15/1M outglm-5.1 — $1.64/1M in, $5.15/1M outglm-5 — $1.17/1M in, $3.74/1M outglm-5-turbo — $1.22/1M in, $4.53/1M outglm-4.7 — $0.70/1M in, $2.57/1M outglm-4.6 — $0.70/1M in, $2.57/1M outglm-4.5 — $0.70/1M in, $2.57/1M outglm-4.5-x — $2.57/1M in, $10.41/1M outglm-4.5-air — $0.23/1M in, $1.29/1M outglm-4.5-airx — $1.29/1M in, $5.27/1M outglm-4-32b-0414-128k — $0.12/1M in, $0.12/1M outglm-5v-turbo — $1.22/1M in, $4.53/1M outglm-4.6v — $0.35/1M in, $1.05/1M outglm-4.6v-flashx — $0.05/1M in, $0.47/1M outglm-4.5v — $0.70/1M in, $2.11/1M outkimi-k2.7-code — $1.11/1M in, $4.68/1M outkimi-k2.7-code-highspeed — $2.22/1M in, $9.36/1M outkimi-k2.6 — $1.11/1M in, $4.68/1M outMiniMax-M2.7 — $0.35/1M in, $1.40/1M outMiniMax-M2.7-highspeed — $0.70/1M in, $2.81/1M outMiniMax-M2.5 — $0.35/1M in, $1.40/1M outMiniMax-M2.5-highspeed — $0.70/1M in, $2.81/1M outMiniMax-M2.1 — $0.35/1M in, $1.40/1M outMiniMax-M2.1-highspeed — $0.70/1M in, $2.81/1M outMiniMax-M2 — $0.35/1M in, $1.40/1M outdeepseek-v4-pro — $0.77/1M in, $2.32/1M outkimi-k3 — $3.51/1M in, $17.55/1M outglm-5.3 — $1.64/1M in, $5.15/1M outminimax-m3 — $0.35/1M in, $1.40/1M outtools (+ optional tool_choice) with the request; the answer then carries choices[0].message.tool_calls (finish_reason: "tool_calls"), and the next turn sends the assistant message back with content: null plus one role: "tool" message per call. The tool schema counts as input tokens. Models verified for tool calling are marked "tools": true in /v1/models; details in /skill.md.npx skills add bridgenode-ai/bridgenode-skillpip install bridgenode-llmpip install bridgenodepip install bridgenode-clipip install bridgenode-sdkpip install bridgenode-mcppip install bridgenode-skillnpm i @bridgenode/llmnpm i @bridgenode/mcp