Most API pricing was designed for a human with a credit card. A person signs up, pokes at the docs for an afternoon, and decides. A fourteen-day trial fits that behavior perfectly. It gives the human a window to evaluate, then asks for a decision.
Agents do not behave like that. An agent that monitors disclosure data does not evaluate once and move on. It calls the same endpoints on a schedule, forever, at low volume. A trial is the wrong shape for that workload. That is why our free tier is not a trial at all. It is 50 calls a month, permanently, with no expiry and no card required. This article explains why that shape fits agents, and how our charge-on-success model handles the awkward case where a query runs correctly and returns nothing.
One note before we start. Nothing here is investment advice. It is a description of how a data product is priced and why.
The trial problem, from an agent's point of view
Think about what a trial actually assumes. It assumes there is a single evaluation phase, followed by a purchase decision, followed by steady production use. That lifecycle describes a procurement process run by a person.
An agent's lifecycle looks different. Someone wires a disclosure-monitoring step into a larger workflow. The workflow runs daily or weekly. Most days it finds nothing worth acting on. The step might run for months at a few calls per day before anyone decides whether it deserves a budget. If the underlying API key dies on day fifteen, the workflow silently breaks, usually long after the person who set it up stopped watching it.
Trials also fail agents in a quieter way. Agents get rebuilt. A workflow gets refactored, redeployed, or handed to a different orchestration framework. Each rebuild needs a working key to test against. A trial that expired three months ago means every rebuild starts with a human doing signup paperwork. A permanent free floor means the agent, or the person maintaining it, can always verify that the integration still works before deciding to spend anything.
There is a third failure mode worth naming. Trials with generous limits invite a burst of exploration, then a cliff. Agents do the opposite of bursting. They trickle. A workload of one or two calls a day never stresses a trial's limits and never survives its expiry. The pricing model and the usage pattern are simply mismatched.
What 50 calls a month actually covers
Fifty calls a month is not a marketing number picked to feel generous. It is sized for the trickle workload described above.
The arithmetic is straightforward. A daily check runs about 30 times a month. That fits inside the free tier with room left for retries, backfills, and the occasional ad-hoc question. A weekly digest uses four or five calls. An agent that checks congressional trade filings every weekday and pulls a 13F consensus snapshot once a week stays under the cap comfortably.
The cap matters as much as the floor. Fifty calls is enough to run a real monitoring loop indefinitely. It is not enough to build a competing data product on top of ours, or to hammer the API with backtesting sweeps. That boundary is deliberate. The free tier exists so that agents can run, not so that heavy workloads can hide from a bill.
There is also a timing reason why low-frequency polling is the honest design for this data. Congressional trade disclosures arrive on a legal deadline measured in weeks, not minutes. We cover the 45-day filing window and its consequences in our piece on the 45-day rule, and the full pipeline from a member's trade to a published document in our walkthrough of how congressional disclosures work. Institutional holdings data has its own quarterly rhythm and lag, covered in our piece on 13F deadlines. The short version is that this data updates slowly by law. An agent that polls it every minute is wasting calls on a feed that cannot move that fast. Fifty well-timed calls a month is not a limitation for this domain. It is roughly the natural request rate.
Charge-on-success, and the query that returns nothing
Above the free tier, we charge per successful result, not per request. This is the part that needs honest explanation, because "success" is a slippery word in a query API.
The mechanics are simple on the happy path. You ask for scored congressional trades matching a filter. We return a result set. You are charged for that result set. If our side fails, times out, or errors, you are not charged. Failed compute is our cost, not yours.
The hard case is the empty result. Suppose your agent asks for new filings by a specific member since yesterday, and there are none. The query executed correctly. We searched the data. The true answer is "nothing new." Is that a success?
Under our model, no charge is applied when a query returns zero results. You pay when we deliver data, not when we deliver an absence of data. We think this is the right default for monitoring workloads, and it is worth being clear about why, including the argument against it.
The argument against is real. An empty result is often valuable information. "No new filings" is exactly what a monitoring agent wants to confirm, and confirming it costs us the same compute as returning ten rows. A strict economist would say we should charge for the answer, not the rows.
We chose not to, for two reasons. First, incentives. If empty results were billable, we would profit from vague queries and from polling that is more frequent than the data warrants. That is a bad position for a data vendor to be in. Charging only on delivered results means our revenue depends on the data actually containing something, which keeps our incentives pointed at coverage and freshness rather than at your query volume. Second, predictability. An agent author cannot always predict hit rates. A filter that matches often in earnings season may match nothing in August. Billing that swings with the news cycle is billing an agent cannot budget for. Billing per delivered result set is at least proportional to the value received.
The practical consequence for agent design: you can poll on a sensible schedule without treating every empty check as a sunk cost. The expensive mistake in most metered APIs, paying repeatedly to learn that nothing happened, is not a mistake here.
The free alternatives, named plainly
All of the underlying data is public and free at the source. That is not a caveat buried in a footer. It is the foundation of the whole category.
The SEC publishes structured, documented, free endpoints for filings and financial facts on its EDGAR application programming interfaces page. Congressional financial disclosures for House members are searchable for free through the Clerk of the House financial disclosure site. What the underlying disclosure law requires of members is covered in our summary of what the STOCK Act actually requires. If you have the time to build parsers, handle amendments and PDF-only filings, normalize names and tickers, and maintain all of it, you do not need us. Plenty of engineers make exactly that choice, and for a one-off research project it is often the right one.
What we sell is not access to public data. It is the normalization, deduplication, and scoring layered on top, delivered as one call instead of a pipeline you maintain. The free tier exists so an agent can test whether that layer is worth paying for, using real production queries, for as long as the evaluation honestly takes. For filings with their own reading quirks, like private placements, our guide to reading a Form D filing shows what that normalization has to deal with.
What this means if you are building an agent
If you are wiring disclosure data into an agent today, the checklist is short. Pick a polling schedule that matches the legal cadence of the data, not the cadence of your cron habit. Budget on delivered results, not on requests, because empty checks are free. And treat the free tier as a permanent floor for integration tests, so every rebuild of your workflow can verify itself without a human in the loop.
If you want to see what a scored result set actually looks like before writing a line of code, start with the report the free tier was sized around. The Congress Stock Trades report shows the same scored disclosure data the API returns, updated on the same schedule your agent would poll. Look at it first, then decide whether the 50 free calls are worth pointing your agent at.
Want the signal instead of the raw filings? Get a free report preview. Prefer the tool to the write-up? Browse all data feeds or connect the free MCP server.