What is AIsa?
AIsa is the resource and payment layer for AI agents: one account and one balance to discover, call, and pay for live LLM and media models and 100+ data APIs - search, finance, social, SEO, sales, and more. See How It Works.How is AIsa different from an LLM gateway?
An LLM gateway puts several models behind one API. AIsa does that, and adds the data APIs agents act on, an MCP server and Skills so agents can find and call them, and per-call billing with spend caps so agents can pay for what they use. If your workload only needs one model provider, a direct integration can be simpler; AIsa helps once an agent needs several providers.Which way should I connect?
- An agent (Claude Code, Codex, Cursor, …): Connect via MCP. OAuth in the browser, no key to paste.
- A coding agent that should learn AIsa’s workflow: add the Agent Skill on top.
- A terminal, script, or CI job without MCP: Use the CLI.
- Your own code: an API key over HTTP or your existing SDK.
Do I need to change my OpenAI or Anthropic code?
No. Change the base URL and the key:https://api.aisa.one/v1 for the OpenAI SDK, https://api.aisa.one for the Anthropic SDK. Standard parameters and streaming work as before. See OpenAI / Anthropic SDK.
What tools does the MCP server expose?
Five:search and list_categories to find an operation, get_details for its contract and a quote, and use or batch_use to run it. The first three are free; use and batch_use are billed per call. Your agent reaches the whole data catalog through these five instead of loading hundreds of tools.
Should my agent use a Skill or call APIs directly?
Use a Skill when an existing one already covers the outcome - it encodes which tools to call, in what order, and how to present the result. Call APIs directly (over HTTP or MCPuse) when your application needs its own sequencing, retries, or data processing. You can combine both: a Skill for the common path, direct calls for the rest.
How is usage billed?
Models are billed per token; data APIs per request, either a fixed rate per call or metered from what the request returns. Everything draws on one prepaid balance, and discovery - docs,/v1/models, MCP search and get_details - is free. The docs explain billing rules but do not list prices; see aisa.one/models and aisa.one/api, and Billing & Credits for quotes and spend caps.
Can I cap what an agent spends?
Yes. Give each agent its own API key with a spending limit in the Console. That limit is the hard ceiling on what the key can spend. For a single call, endpoints that declare maximum-price support acceptX-AISA-Max-Price-USD over HTTP, and MCP use and batch_use accept max_price_usd. If the cost estimate exceeds your maximum, the gateway rejects the request with 402 before calling the provider. This checks the estimate; it does not cap the final charge. Model routes such as /v1/chat/completions do not apply this header, so bound their cost with max_tokens and the key’s spending limit. See Billing & Credits.
Does every endpoint support x402 payments?
No. Only endpoints that declare support accept the x402 machine-payment protocol; check the endpoint’s reference page. See Machine payments for agents.Which models are available?
GPT, Claude, Gemini, Grok, DeepSeek, Qwen and Wan, Kimi, MiniMax, GLM, ByteDance Seed and Seedream, Hunyuan, MiMo, and more, plus embedding, image, and video models. The catalog changes often:GET https://api.aisa.one/v1/models and aisa.one/models are always current, and the model catalog lists IDs and routes.