Plain-English request to collector code
Describe the data you want in plain English and Prometheus generates a TypeScript collector that uses the Firecrawl SDK.
Prometheus
Prometheus is an experimental Firecrawl agent that turns plain-English requests into verified TypeScript collectors, with versioned, self-healing scripts.
Overview
Prometheus is an experimental Forward Deployed Agent from Firecrawl that turns plain-English data requests into verified Firecrawl collectors. It is positioned for people who want reproducible collection code rather than a one-off answer.
A build request returns a TypeScript collector, a sample of the data it produced, and a verified output shape. You can keep that code as-is, save it as a versioned Script, or deploy it so it runs on a schedule or on demand.
The product is designed around a workflow where the agent uses Firecrawl against the live web, writes deterministic code, and checks it before handing it back. If the site changes later, Prometheus can self-heal the collector and append a new version so deployments stay current.
The provided settings and pricing pages also show that Prometheus runs on a connected Firecrawl team account, supports Claude and Codex model engines, and uses token-based access for the CLI and MCP server. Firecrawl pricing includes a free tier, paid monthly plans, and custom Scale and Enterprise options.
Capabilities
Verified output before handoff
Prometheus runs the generated collector before returning it, so the code has already been verified against the live site.
Code and sample data output
A build returns script.ts plus a sample of the data it produced, which you can embed or run yourself.
Versioned, self-healing Scripts
Scripts are versioned and self-healing; when a target site changes, a successful repair appends a new version.
Scheduled and on-demand delivery
Deployments can run on a schedule, on demand, or both, and can serve fresh data as an API endpoint.
Live-web collection workflow
Prometheus can use search, scrape, map, crawl, and interact as part of its headless workflow on the live web.
Use Cases
Create a reusable collector from a prompt
Describe the data you need in plain English and get back code you can inspect, version, and run in your own environment.
Maintain a collector over time
Keep a collector current as a target site changes by saving it as a Script and letting Prometheus repair and version it when needed.
Scheduled data refreshes
Run a Deployment on a cron schedule when you want a predictable refresh cadence for data you rely on regularly.
On-demand fresh data
Trigger a Deployment on demand when you need fresh results immediately and want the response returned inline.
Fit into different development workflows
Use the CLI, HTTP API, MCP tools, or an installable Skill depending on whether you are building in code, scripts, or an agent workflow.
Pros and Cons
Pros
- Turns a plain-English request into reproducible collector code.
- Returns a verified script and sample output before handoff.
- Supports versioned Scripts with self-healing when target sites change.
- Can run as a scheduled deployment, an on-demand run, or both.
- Uses a Firecrawl team account with OAuth rather than requiring pasted API keys for the app connection.
Cons
- The available pages do not provide a full list of integrations or output destinations.
- Prometheus is described as experimental, so some behavior may still change.
- The docs do not show every setup detail in the provided evidence, such as full workflow configuration or limitations.
FAQ
What does Prometheus do?
Prometheus turns a plain-English data request into a verified Firecrawl collector. It builds TypeScript code, runs it to confirm the output, and can save it as a versioned Script or deploy it on a schedule or on demand.
What is the difference between a build, a Script, and a Deployment?
Prometheus supports builds, Scripts, and Deployments. A build is a one-shot request that returns code and sample data, Scripts add versioning and self-healing, and Deployments run that script on a schedule, on demand, or both.
How is Prometheus connected and how long do runs take?
Prometheus uses your Firecrawl team account through OAuth on the settings page. The docs also say builds and prompt-based script creation run the agent and take about 30–180 seconds, while an on-demand deployment run can take up to about 2 minutes.
What output does Prometheus return?
The source describes outputs in terms of a TypeScript collector script.ts and a sample of the data it produced. It also says on-demand deployment runs can return fresh data inline.
What integrations or destinations are supported?
The docs emphasize that Prometheus can use Firecrawl to search, scrape, map, crawl, and interact with live sites. It does not describe a broad integration catalog on the pages provided, so destination and integration details are limited in the available evidence.
Quick Facts
- Category
- Developer Tool
- Product type
- Experimental Forward Deployed Agent
- Primary output
- TypeScript collector plus sample data
- Workflow
- Build, Script, Deployment
- Account model
- Connected Firecrawl team account via OAuth
- Pricing signal
- Free tier, paid plans, and custom Scale/Enterprise options on Firecrawl
Prometheus Alternatives
Happenstance
Happenstance is an AI-powered network search tool for finding people, mutual connections, and warm introductions across connected accounts. It supports individual use, shared team groups, and developer workflows through API, MCP, Slack, and other integrations.
Geekflare Web Scraping API
Geekflare Web Scraping API is a developer-focused web scraping service that extracts content from dynamic pages and returns Markdown, HTML, JSON, or plain text. It is aimed at teams that need browser rendering, CAPTCHA handling, and proxy support without building that infrastructure themselves.
Claro
Claro Research Agent automates manual research in a table-based workflow for list enrichment, company research, document extraction, and pricing monitoring. It can run on its own or connect to the broader Claro platform for entity-aware, system-synced outputs.
Ghost
Ghost is a terminal-based AI assistant for chatting, code generation, and CLI tasks. Includes free models, supports Linux, macOS, Windows, and is open source.
Skayle
Skayle is a content and AI search visibility platform that researches topics before writing, publishes structured content to a CMS, and tracks whether brands are cited in AI search. It is aimed at teams that want one system for publishing, schema-rich content, and visibility monitoring.
SlimSnap
SlimSnap is a Mac app that turns screenshots into JSON for CLI coding agents such as Claude Code, Aider, and Codex CLI. It keeps capture and OCR local, and the site says the app is free with no registration required.