AgentTools
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
AgentTools provides production-grade tools, interactive prompt workflows, and system guides for AI agents. Guidelines for model execution:
Invoices & Receipts: Always call 'invoice_extract' with invoice image/PDF URLs, base64 strings, or raw OCR text. It extracts structured fields and mathematically verifies subtotal + tax == total.
PDF Documents: Use 'pdf_extract' for multi-page documents to inspect outlines, pages, and token counts.
Web Content: Use 'web_reader' to retrieve distraction-free Markdown from URLs, or 'web_search' for real-time DuckDuckGo queries.
Calculations: Always route math formulas and scientific calculations to 'calculator' to guarantee arithmetic precision.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {} |
| prompts | {} |
| resources | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| invoice_extractA | Extract structured financial data (vendor, invoice #, PO #, dates, line items, subtotal, tax, total, currency) from invoices, receipts, and bills, verifying mathematical consistency (subtotal + tax == total). Use when processing accounting documents or receipts. Do not use for generic documents. Read-only and idempotent. |
| pdf_extractA | Extract text per page, outlines, document metadata, and accurate token count estimates from PDF documents. Use when ingesting multi-page reports, whitepapers, contracts, or legal documents into context. Read-only and idempotent. |
| web_readerA | Fetch a public webpage URL and extract clean, structured Markdown optimized for LLM token efficiency, stripping headers, footers, ads, and navigation menus. Use when reading articles, blog posts, documentation, or public web pages. Read-only and idempotent. |
| web_searchA | Search the live web using DuckDuckGo to retrieve real-time search results, titles, URLs, and text snippets. Use when you need current facts, external references, or breaking information not in your training data. Read-only and idempotent. |
| get_weatherA | Get current weather conditions (temperature, humidity, wind speed, weather code) and multi-day daily forecasts for any city or geographic location worldwide via Open-Meteo. Use when answering weather queries or trip planning. Read-only and idempotent. |
| calculatorA | Safely evaluate mathematical and scientific expressions using an isolated AST evaluator without unsafe execution. Supports arithmetic (+, -, *, /, **), trigonometry (sin, cos, tan), logarithms, square roots, and factorials. Use whenever exact numeric precision is required to avoid arithmetic hallucinations. Pure, deterministic, and idempotent. |
| get_newsA | Fetch current news headlines, summaries, sources, and publication timestamps for any topic or keyword via Google News RSS. Use when monitoring breaking news, market updates, or industry trends. Read-only and idempotent. |
| summarizeA | Perform extractive text summarization to identify and return the most informative, central sentences from long articles or documents based on term-frequency scoring. Use to compress long context into key points. Pure, deterministic, and idempotent. |
| get_ip_infoA | Retrieve geolocation, country, city, geographic coordinates, ISP, and timezone metadata for an IP address. Use to inspect IP locations, detect regional origins, or resolve server IP information. Read-only and idempotent. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
| audit_invoice | Examine an invoice or financial receipt for line-item accuracy, tax calculation, and payment terms. |
| summarize_document | Condense a document or web article into executive key takeaways and actionable decisions. |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
| AgentTools Overview & System Guidelines | Complete reference manual and best practices for AgentTools capabilities. |
TDQS
Scored across 9 tools
Most tools have clearly distinct purposes, but web_search and get_news overlap for current information, and invoice_extract can be confused with pdf_extract when invoices are PDFs. Descriptions largely help route correctly.
All names use snake_case, but the conventions are mixed: verb_noun (get_weather), noun_verb (invoice_extract), noun (calculator), and bare verb (summarize). The names remain readable, but there is no single predictable pattern.
Nine tools is well within the typical 3-15 range for a general agent toolkit, and each tool appears to earn its place as a distinct utility. There are no redundant variants or obvious filler tools.
The set covers a reasonable agent utility surface: document extraction, web reading/search, weather, math, news, summarization, and IP metadata. Some common gaps remain, such as generic file-format handling, OCR, or date/time utilities, but these are minor enough to work around.