Firecrawl Alexandria brings 100+ data providers into ChatGPT and Codex

From secondhand summaries to primary sources

Firecrawl describes a familiar limit of AI agents. Ask ChatGPT about a company's latest quarterly filing and it searches the web, finds a page discussing the filing, and summarizes that page. The result is a secondhand account of the numbers, while the filing itself sits in a provider API the model cannot call.

According to Firecrawl, this means agents do well when the open web has already written up a subject, and poorly when the answer depends on a primary source, such as what a series reported last quarter, what a regulator filed, or what a dataset records.

Firecrawl Alexandria is meant to close that gap. The company says it gives the agent the provider's own record, under agreement with the providers who maintain it. One request finds the source, shows what it holds and what it costs, and pulls the data. Firecrawl says providers are paid under agreement when an agent pulls their data. Coverage spans 100+ official providers, plus Firecrawl's Research, Developer and Government indexes.

Firecrawl also cites an internal evaluation. Across 845 tasks with the same model, the same prompts and blind AI judging, agents using Firecrawl Alexandria scored 21% higher on answer quality than agents using built-in web tools.

How it works in ChatGPT and Codex

The plugin follows three steps: find a tool, check what it needs, then run it. The user simply asks a question and the plugin walks through those steps. Firecrawl gives the example prompt "Find remote software engineering jobs posted in the past week. Include salaries where available and application links." ChatGPT searches for relevant tools, reads the inputs and price of the one it picks, and calls it. The user sees which tool was chosen and what the data cost.

The same pattern covers research papers, company filings, government spending records, podcast transcripts, package registries and real estate. If no provider fits, the plugin falls back to web search and scraping, so a coverage gap does not end the question.

Users can also ask what exists before spending anything, for instance by requesting a list of providers that cover federal contract awards along with the price each charges per call.

In Codex, the Developer Index searches repositories, issues, merged pull requests, READMEs and documentation directly and hands Codex the matching passages. Firecrawl notes that answers about library behavior often sit in a README, an issue thread or a merged fix rather than in a blog post. Its example question is how to add pydantic with uv instead of pip. Firecrawl says Codex is where the Developer Index gets the most use.

Billing, terms and setup

Firecrawl states that discovery is free and execution is billed. Asking what data exists, reading a tool's inputs and checking its price cost nothing. Users pay the tool's listed price when it runs. Web search and page scraping are billed as normal.

Some providers require terms to be accepted first. Accepting a provider's terms binds the whole organization, so by default an admin accepts them in the Firecrawl dashboard. When a provider needs this, ChatGPT shows the terms and points to the page. If an organization has enabled agent acceptance, ChatGPT may accept them after showing the current terms.

A Firecrawl account is required. Alexandria is not available to keyless sessions, so the plugin must be connected to an account before ChatGPT can discover or run any provider tools. Users who already have the official ChatGPT plugin installed will see Firecrawl Alexandria in their next conversation. Others can open ChatGPT, install the Firecrawl plugin and connect their Firecrawl account when prompted. The same plugin covers both ChatGPT and Codex.

What it means for users

For people who rely on ChatGPT or Codex for research, the change is that answers can come from a provider's structured data rather than from a web page that quotes it. Because prices are shown before a tool runs and discovery costs nothing, users can see what a query will cost before committing. Developers using Codex gain a searchable source of repository and issue material for questions about how libraries behave in practice.