Introduction

We are pleased to announce a new data partnership between CityFALCON and Proxygate. As of today, CityFALCON’s API is live on Proxygate, a marketplace where AI agents discover and call APIs directly. Agents and the developers building them can now access CityFALCON’s real-time financial news, sentiment, insider transactions, investor relations content, and official filings, priced per request rather than monthly.

For developers and teams building financial agents, this is a meaningful shift in how financial data can be consumed. Instead of negotiating enterprise contracts with fixed terms, agents can now discover the endpoint and call exactly the data they need.

This article explains what the partnership means in practice, why pay-per-call financial intelligence is particularly important for AI agents, and what makes pre-processed financial data fundamentally different from running raw news through a large language model.

 

What’s Available on Proxygate

Through the new listing, agents and developers can access the full breadth of CityFALCON’s financial intelligence:

  • Real-time news from over 10,000 sources in more than 50 languages, with relevance scoring via the CityFALCON Score
  • Sentiment at the story level and as a timeseries across assets and topics
  • Insider transactions, investor relations content, and official company filings, structured and query-ready
  • All delivered as pre-aggregated, scored, NLU-tagged data, accessible via standard API

This is the same structured data infrastructure trusted by eToro, BNP Paribas, StockTwits, and StoneX, now available through Proxygate’s agent-first marketplace.

 

Why Pay-Per-Call Matters for AI Agents

Traditional enterprise data contracts were not designed with autonomous AI agents in mind. They assume a procurement process, a sales call, a contract negotiation, and a fixed monthly fee that covers expected usage. None of this fits how agents actually work.

Agents are designed to take actions in real time. They cannot wait for a contract to be signed or amended for new use cases and datasets. They need data that is callable, structured, and priced per request. They need to discover endpoints autonomously, evaluate whether the data fits their query, and pay only for what they actually consume.

Proxygate is built for this model. As a marketplace for the agent economy, it lets agents discover and call APIs directly. The CityFALCON listing brings pre-processed financial intelligence into this model, so agents have access to high-quality structured data without the cost overhead of an enterprise contract, the aggregation infrastructure, or running irrelevant text through an LLM.

 

The Hidden Cost of Building With Raw LLMs

To understand why pre-processing infrastructure matters for agents, it helps to look at the alternative. Large language models can analyse financial news directly, but the economics break quickly at scale.

A single news item might be a few hundred tokens. Add a system prompt with instructions, and a request can easily run a couple thousand tokens just for input. The output adds more. Multiply that across hundreds or thousands of stories per day, and even fractions of a cent per token translate into bills running into thousands of dollars per day just to process the data, before you get any insights.

For an agent that needs to monitor news for a portfolio of companies, react to events in real time, and make decisions based on what it learns, this cost structure is unsustainable. Every query consumes tokens, and every action requires multiple queries. The agent ends up spending most of its budget extracting information that was already known and structured by someone else.

And this does not even account for the cost of web search and the entire infrastructure that goes into surfacing this content, the problem of overwhelming models with context from very similar or even identical stories in the pipeline, and the inefficiency of processing everything that comes into a news pipeline, even irrelevant content that is simply thrown away.

 

A Real-World Example: The Cost of Monitoring the Market

Consider an agent monitoring news for 50 publicly traded companies in real time. A typical universe of 50 companies may generate a few hundred to a couple thousand stories per day. Many of these are redundant, some are irrelevant, and there is no structure.

This may add to tens or even hundreds of dollars per day, just for processing input. Add output costs, retries, quality control, and the supporting infrastructure, and your monthly bill quickly runs into the thousands or tens of thousands of dollars range. 

Now expand the coverage. Tracking hundreds of companies multiplies the volume. Adding global coverage in dozens of languages multiplies it again. 

For an autonomous agent that needs to operate at this scale, which could be pretty common for most individual’s portfolios, and almost certain for any larger entity, the token-cost math simply doesn’t work.

 

Where LLMs Genuinely Excel

This isn’t to say LLMs aren’t useful. They are extraordinary tools and have a clear place in financial intelligence. They excel at:

  • One-off analysis. Summarising a complex document, extracting structured data from a single filing, or answering a specific research question.
  • Repetitive analysis of curated content. Updating positions and outlooks based on the latest information that the LLM can assume to be already-vetted and relevant.
  • Custom or unusual tasks. When you need flexibility that pre-built systems don’t offer, LLMs let you describe what you want in natural language.
  • End-user interaction. Chatbots, research assistants, and natural language interfaces benefit from conversational flexibility.
  • Low-volume, high-value work. Investment memos, deep dives on specific companies, or complex research where the cost per query is justified by the importance of the answer.

What LLMs don’t do well, at least not cost-effectively, is broad, continuous, structured processing of large volumes of repetitive, repeated data. That’s where pre-processed infrastructure wins.

 

The Infrastructure Approach

CityFALCON’s approach is to do the processing at scale and make the results available as structured data. Instead of paying per token to process each article from scratch, you access content that has already been aggregated, scored, tagged, and indexed.

Our infrastructure already:

  • Aggregates content from over 10,000 financial sources across more than 50 languages
  • Scores every story for relevance to specific tickers, sectors, and topics using the CityFALCON Score
  • Tags stories with sentiment, NLU entities, and proprietary scoring
  • Provides insider transactions, investor relations content, and official filings in the same API
  • Filters out duplicates and noise
  • Delivers everything via a structured API

The structured output means agents don’t need to spend tokens figuring out what’s in each story. You pay for a data stream of pre-processed content and can save token costs for the meatier analysis that LLMs do well.

 

Use Cases Where Per-Call Infrastructure Wins

Several common AI agent and developer use cases are particularly well-suited to this model:

Trading agents that react to news in real time can call CityFALCON’s sentiment scores and filter on relevance scores directly, without spending tokens parsing raw articles.

Research agents that scan thousands of companies for opportunities can query structured data at low cost, without budget pressure from per-token processing.

Compliance and risk agents that monitor news for regulatory or reputational events across a broad universe need consistent, structured coverage at predictable cost.

Backtesting workflows that need years of historical sentiment and event data can run efficiently against pre-computed datasets rather than paying to process millions of historical articles through an LLM (editor’s note: this is available directly via CityFALCON, not on Proxygate)

Custom workflows built by developers can mix and match calls to news, sentiment, insider transactions, IR content, and filings, paying only for the specific data their workflow needs.

In each case, pay-per-call infrastructure delivers better economics, more reliable performance, and faster development cycles than building on raw LLMs alone.

 

Conclusion

The agent economy is changing how financial software gets built. Agents need data that’s structured, callable, and priced for how they actually consume it. CityFALCON’s listing on Proxygate brings exactly that: real-time financial intelligence, aggregated and scored, available on a per-call basis with no contract.

Whether you’re building a trading agent, a research tool, a compliance system, or anything in between, this partnership makes high-quality financial data directly accessible without the friction of traditional procurement or the runaway costs of raw LLM infrastructure and processing.

Explore the CityFALCON listing on Proxygate: https://proxygate.ai/seller/cityfalcon/cityfalcon-api