Claude Sonnet 5 API and MCP for AI agents
TollNode gives Claude, Cursor or any AI agent Claude Sonnet 5 as a tool (write, review, summarize, translate or reason about text), paid per call with no API key, from $0.034 per call.
Ask in plain words, like 'review this contract clause and list anything one-sided', and get the written answer back.
No subscription · Sign-up optional · From $0.034 per call · Charged only when it works · No Anthropic account needed
Review this payment clause and list anything one-sided: prices can change without notice, and invoices can only be disputed within 7 days.
TollNode · Anthropic · quick$0.034 · 1 answer · quick
| Issue | Why it matters |
|---|---|
| No price change notice | Prices can rise with no warning, so budgets can't be planned |
| No cap or justification | No limit on how much or how often prices change |
| Very short dispute window | 7 days is short to review invoices and find errors |
| Total waiver of dispute rights | Even billing errors or fraud can't be challenged after 7 days |
Real result, shortened.
What your agent gets from a call
Prompt and instructions
Claude Sonnet 5 API for AI agents
Say what to do, and paste the text it needs.
- Prompt
- Review this contract clause and list anything one-sided: 'The Supplier may change prices at any time without notice.'
- Length
- quick
- System
- optional, e.g. 'You are a contracts lawyer.'
- Returns
- text
You get
- prompt: what you want done, with any text it needs
- system: optional instructions that frame every answer
- text: the answer, returned as plain text
- length: quick, standard or long; sets the price
- limits: each option caps text in and text out
- from $0.034
Length and price
Pick how much text goes in and comes out.
- quick
- 10,000 characters in, 750 words out
- standard
- 40,000 characters in, 3,000 words out
- long
- 140,000 characters in, 6,000 words out
You get
- quick: about 10,000 characters in, about 750 words out
- standard: about 40,000 characters in, about 3,000 words out
- long: about 140,000 characters in, about 6,000 words out
- default: quick
- charged: only when an answer comes back
Every result is structured JSON your agent can sort, filter, summarize or turn into a spreadsheet.
What people ask their AI to do with it
Operations teams
Summarize a long thread
“Summarize this 40-message thread and list the decisions and open questions”
standard
Cost today$0.133
Sales teams
Draft a follow-up
“Write a short follow-up email after this call transcript, friendly and specific”
quick
Cost today$0.034
Contract reviewers
Flag one-sided terms
“Review this clause and list anything one-sided; I am not asking for legal advice, just a first pass”
quick
Cost today$0.034
Support teams
Translate a reply
“Translate this support reply into Spanish and keep the tone polite”
quick
Cost today$0.034
Analysts
Read a long report
“Summarize this 60-page report in 10 bullet points and list the risks it names”
long
Cost today$0.330
Developers
Review a change
“Review this diff and list bugs and unclear names”
standard
Cost today$0.133
Use it as a Claude Sonnet 5 MCP server in Claude, Cursor or Codex
Connect once. Then just ask.
Paste this into Claude Code, Cursor or Codex. It connects TollNode and gives your agent a small wallet. Add a few dollars of USDC and you're done.
npx agentcash add https://tollnode.comCopy for your AI agent
Paste this into your agent so it knows what this node does and how to call it.
You can use TollNode's Claude Sonnet 5 node. Call POST https://tollnode.com/api/v1/nodes/claude-sonnet-5 with length (optional), prompt (required), system (optional). The price depends on length (quick, standard, long; default quick) and is charged only when a result is returned. Over MCP, call the "claude-sonnet-5" tool; use list_models to see prices. Return results as tables when I ask for lists.Price
from $0.034
per call
The price depends on the length you choose. If you don't choose, you get quick.
| quickdefault | $0.034 per call |
| standard | $0.133 per call |
| long | $0.330 per call |
Charged only when a call returns a result. Failed or empty calls cost $0.
Claude Sonnet 5 without an Anthropic account
No subscription, no credit packs
Pay per clip: $0.034 up to quick pages, $0.133 up to standard pages, $0.330 up to long pages.
No Anthropic account, no API key, no billing setup
Your agent calls TollNode. You never set up billing or request access with anyone else.
Only charged when it works
If a call returns nothing or fails, it costs nothing.
One connection, much more
The same setup also gives your agent Google search, YouTube, Maps, images and video, so one prompt can chain them.
FAQ
- Is this the official Anthropic API?
- No. TollNode is an independent service. It gives your agent access to Claude Sonnet 5 through its own API and MCP server, paid per call, with no Anthropic account.
- Can I use it as a Claude Sonnet 5 MCP server in Claude?
- Yes. Add https://tollnode.com/mcp in Claude Code, Claude Desktop or Cursor, and the "claude-sonnet-5" tool appears with its fields and price.
- How much text can I send?
- It depends on the length option. quick takes about 10,000 characters and writes about 750 words, standard about 40,000 and 3,000 words, long about 140,000 and 6,000 words.
- Does it remember earlier calls?
- No. Each call is independent, so include any earlier text in the prompt. Use system for instructions you want applied every time.
- Is it legal advice?
- No. A clause review is a first pass to spot terms to question. Have a lawyer read anything that matters.
- What if nothing comes back?
- You're not charged. TollNode only collects payment when a usable result is returned, and your agent can try a different input.
- How do I pay for Claude Sonnet 5?
- Your agent sends the request, gets an HTTP 402 with the exact price (from $0.034 per call), signs a USDC payment on Base and retries. Or use an account balance with an API key. Either way you are charged only when a result comes back.
- What does the Claude Sonnet 5 response look like?
- A JSON object with the answer in a text field. Each call is independent, so include any earlier text in the prompt.
Works well with
- Document to Markdown (PDF, Word, slides)Turn a PDF into Markdown, then ask Claude about itfrom $0.007
- URL to MarkdownRead a web page, then ask Claude to summarize it$0.006
- YouTube transcriptGet a video transcript, then ask for the key points$0.005
- Speech to text: transcribe an MP3 or WAVTranscribe a recording, then ask for action itemsfrom $0.015
Ready in a minute
Connect your AIFor developers: endpoint, inputs and an example response
POST https://tollnode.com/api/v1/nodes/claude-sonnet-5
Input
| Field | Type | Notes |
|---|---|---|
| length | enum | quick: up to ~3,000 tokens in (about 10,000 characters) and ~750 words out. standard: ~12,000 in (40,000 characters) and ~3,000 words out. long: ~40,000 in (140,000 characters) and ~6,000 words out. Default quick (quick, standard, long) |
| prompt* | string | What you want Claude to do, with any text it needs |
| system | string | Optional instructions that frame every answer, e.g. 'You are a contracts lawyer. Answer in plain English.' |
Example response
{
"text": "This clause is one-sided in three ways: …"
}Also described in the OpenAPI file and llms.txt. Setup for every app: Agents & MCP.