Giving every email in your CSV its own private investigator

What it does

Fire Enrich ingests a CSV of email addresses, extracts the domains, and dispatches a sequence of specialized agents to investigate each company. The system hunts down industry classifications, funding rounds, tech stacks, and leadership details, then synthesizes the findings into a structured table with citations. It is essentially a web-scraping research assistant built on Firecrawl and OpenAI’s GPT-4o, wrapped in a Next.js interface.

The interesting bit

Instead of firing off a single prompt and hoping for the best, the project runs a five-phase assembly line—Discovery, Profile, Financial, Tech Stack, and General Purpose—where each agent builds on the last one’s notes. This sequential handoff is the core gimmick: the funding agent uses the industry context found by the profile agent to search the right databases, and a final synthesis layer resolves conflicts before anything reaches the user.

Key highlights

• Specialized agents: Each phase has a dedicated agent with its own Zod schema, search strategy, and expected output (e.g., FundingResult, ProfileResult).
• Parallel within, sequential across: Every agent fires multiple Firecrawl searches at once, but the phases themselves run in order to share context and validate prior results.
• Cited sources: Every data point is traced back to a URL—Crunchbase, TechCrunch, GitHub, or the company’s own site.
• Extensible schemas: Adding a new data field means editing a Zod schema and updating the orchestrator’s routing logic; the General Purpose agent catches anything unmapped.
• Real-time UI: Results populate live in a Next.js table as the multi-agent pipeline finishes each row.

Caveats

• Hard API dependency: The tool requires active, paid API keys for both Firecrawl and OpenAI to function at all.
• Unclear coverage: The examples and architecture diagram center on corporate domains (the diagram explicitly flags “Corporate email detected”); the README never addresses what happens with personal or public email providers.
• Opaque economics: Each enrichment triggers numerous external API calls across five phases plus a final GPT-4o synthesis, yet the README provides no guidance on cost, rate limits, or caching.

Verdict

Sales teams and recruiters tired of manually researching prospects will find this a ready-made force multiplier, but anyone looking for a cheap, offline, or fully self-hosted enrichment pipeline should keep scrolling.